EA007776B1 - Content conversion method and apparatus - Google Patents

Abstract

A method and apparatus for analyzing documents and thereby determining the association between words in a language. The method includes providing a collection of documents, selecting a first word or word strings, and a second word or word string occurring in the documents. The method further involves associating first word or word strings and second word or word strings with common word or word string based on frequency of occurrence of the common word or word strings within the ranges.

Description

Related Applications

This application is a partial continuation of US Patent Application No. 10/024473, filed December 21, 2001, and claims priority on the basis of US Provisional Application No. 60/276107, filed March 16, 2001, and provisional application No. 60/299472 US patent filed June 21, 2001, each of which is incorporated into this application by reference.

Computer Program Listing Application

This application contains the following computer program listings in the computer program listing application on DISC 1: Rgodgat1 in the rgodgat 1.1x1 file created on May 28, 2002. and containing 12 kB; Rgodgat2 in the rgodgat2.1x1 file created on May 28, 2002 and containing 2 kbytes, Rgodgat3 in the rgodgat3.1x1 file created on May 28, 2002. and containing 17 kbytes, and Rgodgat4 in the file rgodgat4.1x1, created on May 31, 2002. and containing 8 kb.

FIELD OF THE INVENTION

The present invention relates to a method and apparatus for creating a database for use in converting a language from one state to another state.

BACKGROUND OF THE INVENTION

Known devices and methods for automatically translating documents from one language to another language. However, these devices and methods often give inaccurate translation of documents from one language into another language, work slowly and may be inconvenient for use. In addition to human-assisted translation tools, other known devices include widely available machine translation software. These known systems have disadvantages due to which these systems are error-sensitive, have a low speed and are inconvenient. Known devices and methods of translation are not able to constantly produce accurate translations of input texts and therefore often require the user to actively participate in verification reading and editing. Accurate machine translation is more complicated than that produced by devices and methods that perform interlinear translation of documents. The mentioned systems of interlinear translation often give translations that are not well understood by readers of the translated document, since the method of interlinear translation does not provide the correct choice of words and leads to meaningless grammatical units.

To get rid of the described shortcomings, the known translation devices for decades aimed at the choice of word translations in the context of sentences based on a combination or set of lexical, morphological, syntactic and semantic rules. These systems, known in the art as machine translation systems (MT, MT), “rule-based”, have a significant drawback, since there are so many exceptions to the rules that these systems are not able to constantly produce accurate translations.

In addition to the MP-based systems of rules, in the last decade a new MP method, known as MP, "based on examples" (ЕВМТ, ЕВМТ) has been developed. The EUMT method is based on the use of sentences (or, possibly, parts of sentences) stored in two different languages in a multilingual database. If the translation request matches the sentence in the database, then the translation of the sentence into the language into which the translation is done (output language) is performed by the database issuing the exact translation in another language. If the part of the translation request corresponds to the part of the sentence in the database, then the mentioned devices try to determine exactly which part of the sentence converted to the sentence in the input language is the translation of the request.

ЕВМТ systems are not able to give an accurate translation from a language without restrictions, since databases of interlanguage sentences are created manually and therefore will always be, as a rule, "incomplete". Another drawback of EUMT systems is that translations of partial match options are unreliable. Attempts have been made to automate the creation of multilingual databases of pairs of translated documents for use in EBMT systems. However, these efforts have not led to the creation of useful, accurate multilingual databases of any substantial size. None of the mentioned attempts is associated with the use of an algorithm that reliably and accurately produces translations of a significant number of words or phrases from a pair of translated documents.

Some translation devices combine both MP mechanisms, namely MP based on rules and MP based on examples (EBMT). Although this combination of techniques can improve the degree of accuracy compared with the case of using any one system from these systems, the results remain incomplete without serious support from the user and performing editing.

The problems that arise when translating documents from one language into another language can be attributed to the more general problem of converting data expressing concepts or information in one state, for example, text, into data expressing concepts in another state, for example, in the form of mathematical symbols. In these cases, a search should be made in inter-conceptual associative databases that associatively associate data in one state with equivalent data in another state. Therefore, there is a need for an improved and more efficient method and apparatus for creating dictionaries or databases that associate equivalent concepts in different languages or in different states (for example, in the form of words, a word

- 1 007776 combinations, sounds, movements, etc.), and for translating or converting concepts transmitted by documents in one language or in one state into the same or similar concepts presented by documents in another language or in another state.

The invention relates to the processing of content (meaningful information content) using an inter-conceptual associative database. In particular, in accordance with the present invention, there is provided a method and apparatus for creating a database of associated concepts, as well as a method and apparatus for using said database in order to transform concepts from one state to other states.

In accordance with one embodiments of the present invention, there is provided, for example, a method and apparatus for creating a database for translation from one language to another, wherein two languages constitute a database of associated concepts. The present invention also provides a method and apparatus for using said language database in order to convert documents (reflecting concepts) from one language form to another language form (or, more generally, from one state to another state). However, the present invention is not limited to translations from one language to another language, although the description of this preferred embodiment is provided below. The aspect of the present invention, which consists in creating a database, can be applied to any concepts that are somehow related, but expressed in different states, and the aspect of the present invention, which consists in conversion, can be applied to the exact transformation of concepts from one state to another state.

In accordance with another embodiment of the present invention, the aspect of the present invention, which consists in creating a database, can be used to create such associations between concepts within the same language and their relationships with each other, which are used in artificial intelligence applications.

The following is a description of the application of the present invention in an embodiment of translation from one language into another language. In the context of the present description, the terms related to conversion, translation and processing are used as equivalent, and in their broadest sense.

SUMMARY OF THE INVENTION

The objective of the present invention is to provide an efficient translation of documents from one language or from one state to another language or to another state by developing a method and apparatus for creating and maintaining inter-conceptual associative databases. Said databases, in the general case, associate data presented in the first format or state that reflect specific concepts or pieces of information with data presented in a second format or state that reflect the same specific concepts or pieces of information.

Another objective of the present invention is to provide translation of documents from one language or from one state to another language or to another state by developing a method and apparatus for creating a second document containing data in a second state, format or in a second language, from the first document containing data in the first state, format, or in the first language, with the first and second documents reflecting essentially the same concepts or information.

Another objective of the present invention is to provide translation of documents from one language or from one state to another language or to another state by developing a method and apparatus for creating a second document containing data in a second state, format or in a second language, from the first document, containing data in the first state, format or in the first language, the first and second documents reflect essentially the same concepts or information, and the method and apparatus include using ezhponyatiynoy associative database.

Another objective of the present invention is to create a translation of documents (in the broad sense, in the conversion of concepts from one state to another state) in real time.

These and other objectives of the present invention are solved using the method and device for creating an inter-conceptual database. A method and apparatus for creating an inter-conceptual database may include the creation of at least two documents, each document being written in a language different from the others, but reflecting essentially the same concepts. Documents can be exact translations of the same text, that is, they can be parallel text documents, or they can be translations containing texts related in a general sense, i.e. comparable text documents. In accordance with the present invention, at least the first and second occurrences of all words and phrases are selected, which are characterized by a large number of occurrences in the first language in existing multilingual documents. Then, at least one interval of the first word and one interval of the second word in documents in the second language are selected, while the intervals of the first and second words correspond to the first and second occurrences of the selected word or phrase in documents in the first language. Then the words and phrases are compared,

- 2 007776 found in the interval of the first word, with words and phrases found in the interval of the second word, find the words and phrases common to the intervals of both words, and enter the data found common words and phrases into the inter-database. Then, in accordance with the present invention, common words or phrases found in two intervals in the second language are associated with the selected word or phrase in the first language in the said inter-conceptual database, ordered by the frequency of their association (the number of repeated occurrences), after correcting the correction of frequencies associations as described herein. The number of associations highlighted by the database based on the analysis of parallel or comparable texts for common words or phrases in different languages will increase with the number of available parallel or comparable texts in many different languages.

In addition, these and other objectives of the present invention are achieved using the method and device for converting a document from one state to another state. In accordance with the present invention, a database is created containing data segments in the first language associated with data segments in the second language (created by the above methods or manually). In accordance with the present invention, the text is translated by referring to the aforementioned database and identifying the longest word combination in the translated document, starting with the first word of the document that exists in the database. The system then retrieves the second language phrase from the database associated with the found phrase from the document in the first language. The system then selects the second phrase in the document that exists in the database and contains an overlapping word (or, alternatively, the phrase) with the previously identified phrase in the document, and extracts the second language phrase associated with the second phrase in the first language from the database . If the associated phrases in the second language contain an overlapping word (or, alternatively, words), then the associated phrases in the second language are combined (with the exception of redundancy in overlapping) to create a translation; otherwise, other associations in the second language are extracted to phrases in the first language and they are checked for the possibility of combining word overlap up to a positive result. The next collocation in the document in the first language is selected by recognizing the longest collocation in the database, which contains an overlapping word (or, in another variant, words) with the previously recognized collocation in the first language, and continue the above procedure until the entire document in the first language is converted to a second language document.

In addition, in accordance with the present invention, frequency tables are created to determine the association between at least two words or phrases so that these frequency tables can be used in other applications, including applications related to the conversion of content from one state to another state. Frequency tables are created by analyzing documents in a given state (for example, in a given language) and determining the frequency with which two words and / or phrases are related based on proximity to a word or phrase in the text. Therefore, for example, according to the results of the analysis of texts in English, frequency tables can be set for words or phrases related to the phrase “Everest peak” (Moi! Eyetey), for example, “mountain” (toip1ash), “the highest place in the world” (11f11C51 p1ase ίη 111е \\ όγ16). "Snow" (5po \ y). "Ascent" (sbtb), "dead" (Reb1e b1eb) and "cold" (sbb). Mentioned frequency tables can then be used by any means in any intelligent applications to answer questions by recognizing common associations from at least two frequency tables. Databases created for intelligent applications can be built on documents in one language (or, alternatively, using interlanguage text).

List of drawings

FIG. 1 is an embodiment of an inter-concept database in accordance with the present invention.

FIG. 2 is an embodiment of a computer system in accordance with the present invention in order to implement the methods of the present invention.

FIG. 3 is a memory device (memory) of a computer system in accordance with the present invention for storing programs that implement the methods of the present invention.

DETAILED DESCRIPTION OF THE INVENTION

In accordance with the present invention, there is provided a method and apparatus for creating and maintaining an inter-conceptual database and for translating documents from a first language or from a first state to a second language or into a second state using an inter-conceptual database. The documents considered in the present description are the foundation of such information as concepts represented by symbols and letters fixed on some medium. For example, the documents may be electronic documents recorded on magnetic or optical media, or paper documents, such as books. Symbols and letters contained in documents represent concepts and information expressed using one or more display systems, with respect to which it is assumed that they are understandable to document users. In accordance with

- 3 007776 of the present invention, documents in the first state, i.e. containing information expressed in one display system is processed in order to obtain documents in the second state, i.e. containing essentially the same information expressed using a second display system. Therefore, the present invention makes it possible to process or translate documents from one display system into another display system, for example, written documents or spoken language from languages such as English, Hebrew and Cantonese, into other languages.

The system or apparatus for implementing the content conversion or the content processing method in accordance with the present invention may, respectively, be a computer system 200 shown in FIG. 2, or implemented on the basis of this system. The computer system 200 comprises a processor 202 connected through a bus 214 to a memory 208, an input device 210, and an output device 212. The computer system 200 may also include a memory device 204 and a network interface 206. The processor 202 accesses data and programs stored in memory 208. By executing programs in memory 208, a processor can control a computer system 200 and perform data processing and device management steps, such as input device 210, output device 212, memory 204, and network interface 206 and memory 208. Programs stored in memory 208 may include steps implementing the method in accordance with the present invention, for example, converting content, associating words and phrases and creating a database, and tracking methods.

The memory 204 records and stores information for subsequent retrieval in the memory 208 by the processor 202 and may contain known memory devices, such as non-volatile memory, magnetic disk drives, magnetic tape drives and optical memory. The memory 204 can store programs and data, including databases, which can be transferred to the memory 208 for use by the processor 202. The entire database or database blocks can be transferred to the memory 208 for access and processing by the processor 202. The network interface 206 provides data exchange between the computer system 200 and the network 216, for example, the network 1p1sgps1. and converts the signals coming from the computer system 200 into a format that can be transmitted over the network 216, and vice versa. Input device 210 may comprise, for example, a keyboard and a scanner for inputting data into memory 208 and into memory 204. Input data may comprise text of documents to be written to a document database for analysis and conversion of content. An output device 212 includes devices for presenting information to a user of a computer system and may include, for example, a monitor screen and a printer.

The following is a detailed description of the present invention, including a method and apparatus for creating a database and a method and apparatus for converting.

Method and device for creating a database

In accordance with the method of the present invention, an inter-database database is used to process document content. In FIG. 1 shows an embodiment of an inter-concept database. This embodiment comprises listing segments of associated data in columns 1 and 2. Data segments are characters or groupings of characters that represent a particular concept or piece of information in a display system. If the display system in a document is, for example, a natural language, then the segment may be a word or phrase. Therefore, the segments of system A in column 1 are data segments that represent different concepts and a combination of concepts, Ea1, Ea2, Ea3 and Ea4 in the hypothetical mapping system A. The segments of system B in column 2 are EI, E3, E4, E5, E7, E9, ΌΜ0 and ΌΜ2, which represent various concepts and some combinations of such concepts in a hypothetical mapping system B, which are ordered by the frequency of associations with data segments in mapping system A. Column 3 indicates the direct frequency of associations, which represents th number of cases when a segment or segments in language B were associated with a given segment (or segments) in language A. Column 4 presents the frequencies of associations after subtraction, which are the number of cases when a segment (or segments) of data in language B were associated with a segment (or segments) in language A after subtracting the number of cases where a segment (or segments) were associated with a larger segment, as described below.

As can be seen from FIG. 1, a situation is possible in which a single segment, for example Ea1, is most appropriate to associate with several segments, E1 together with E3 and E4. The higher the association frequencies after subtraction (in accordance with the present description) of the data segments, the higher the likelihood that the segment of system A is equivalent to the segment of system B. In addition to measuring the adjusted frequencies by the total number of occurrences, the corrected frequencies can also be measured, for example, by calculating the percentage of the number of times when specific segments of system A corresponded to specific segments of system B. If the database is used to translate a document, then in accordance with the procedures th associated segment of the highest rank will be retrieved from the database first. However, often the method used to verify the combination of associated segments for translation (see below) leads to the decision that

- 4 007776 Take another association with a lower rank, since a verified association of a higher rank cannot be applied. For example, if the database receives an association request for Ea1. the database will give the answer ΌΜ + Ό63 + Ό64; but if the association ΌΜ + Ό63 + Ό64 cannot be applied in accordance with the decision of the procedure. which correctly combines data segments for translation. then the database will give E9 + E0 to check for the correct combination with the other associated segment for the translation.

Usually. A method for creating an inter-conceptual database in accordance with the present invention comprises the steps of analyzing and working with parallel or comparable text. The method and device in accordance with the present invention are used in this way. to create a database. containing associations of two states - exact transformations or. more specifically. associations between concepts. expressed in one condition. and concepts. expressed in a different state. Translation associations and other important associations between the two states are amplified. those. become more frequent. with an increase in the number of documents. which are subjected to analysis and processing in accordance with the present invention. therefore, as a result of working with a sufficiently large number of “model” documents, the most general (and, in the only sense, accurate) associations are revealed. and the method and device can be used for conversion.

In one embodiment of the present invention, the two conditions are natural languages (e.g., English, Hebrew, Chinese, etc.). and therefore, in accordance with the present invention, a multilingual database is created. bringing words and phrases in one language into line with their translated equivalents in another language. Phrases can be defined as groups of consecutive adjacent words. often with punctuation marks and any other characters. used to express thoughts in language. In this embodiment, in accordance with the present invention, a database is created by analyzing documents in two languages and creating a translation database for each repeating word or phrase in both languages. However, the present invention is not limited to translation from one language into another language. The present invention enables the user to create a database of concepts and associate these concepts with other concepts. with hierarchical distinction of concepts. Consequently. concepts are associated with other concepts and classified by frequency of occurrence. Specific gravity. attached to the frequency of occurrence. and the scope of the database. created by the above method. may vary depending on the needs of the user.

For example. with regard to the translation of text from one language to another language, the effect of the present invention will be to create language translations of words and phrases from English into Chinese and vice versa. In accordance with the present invention, rank ratings of associations between words and phrases in two languages will be issued. If the sample is large enough. that is a word or phrase. often. will be one of the Chinese equivalents of an English word or phrase. However, the present invention will also produce other associations in Chinese for English words or phrases. and the user will be able to work with these associations as he sees fit. For example. in response to the word “mountain” (toyp (shp)), when processed in accordance with the present invention, a list of Chinese words and phrases in the analyzed language can be given. Chinese language equivalents of the word “mountain” (toyp! at). Most likely, the most high rank; however, in accordance with the present invention, other words or phrases in a foreign language will be issued associated with the word "mountain" (toyp (at). For example, "snow" (kpote). "skiing" (kk1). "dangerous sport "(and baidetoik cut)." the highest point on earth "((Not Ydyek! rosh (u (Not \\ όγ16) or “Mount Everest” (Μΐ. Euovey). The user will be able to discretely work with these words and phrases that are likely to be ranked lower than translations of the word “mountain” (toip ( a) In accordance with the foregoing, the present invention is an automated creator of an associative database, the most powerful associations being “translations” or “transformations” in a single sense. and other frequent (but less powerful) associations reflect concepts. which are closely related to the analyzed concept. Therefore, databases can. eg. used by systems well known in the art. using artificial intelligence methods. Currently, these systems use incomplete. manually created databases of concepts or ontologies as "neural networks" for solving applied problems. Mentioned databases of associated concepts for use in artificial intelligence applications can be constructed using any user-defined interval in a document in one language.

In accordance with another embodiment of the present invention, a computing device is used. for example, a computer system based on a widely available personal computer of a known type. Although the computing device is usually a regular personal computer (stand-alone or working in a network environment). the possibility of using other computing devices is also contemplated. e.g. personal digital information devices (ΡΌΑ). wireless devices. servers. large computers, etc. However, the method and device in accordance with the present invention do not need to use the mentioned computing devices and allow the possibility of simple implementation by other means. including the creation of hand

- 5 007776 cross-associations. The method according to which the analysis of sequential documents is carried out in order to expand the collection of “model” documents and creates a cross-associative database is characterized by a variety, namely, documents can be entered for analysis and processing manually, using automatic feeding (for example, automatic paper loaders) or using the search "agents" of the 1Schegpek network, for example, Shel SgaMegs, to automatically select the appropriate documents.

It should be noted that the present invention allows the creation of an associative database by analyzing comparable text, in addition to (or even instead of) parallel text. Further, the method provides for a joint viewing of all available documents when searching for a repeating word or phrase in a language.

Building a database

In accordance with the present invention, documents are analyzed to build a database. After entering the document (again, a pair of documents representing the same text in two different languages), the creation procedure begins using the methods and / or devices specified in the present description.

For illustration, suppose documents have identical content (or, in a general sense, concepts) in two different languages.

Document A is written in language A, document B is written in language B. Documents have the following text:

Document A (language A) Document B (language B) ΧΥΖΧΜνΥΖΧΖ AA BB SS AA EE EE 6C SS

The first step of the present invention is that the word spacing is calculated to determine the approximate location of possible associations to any given word or phrase. Since interlanguage literal analysis alone will not provide a useful result (for example, word 1 in document A often does not exist as a literal translation of word 1 in document B), and a sentence in one language, by its structure, may contain an equivalent concept in another place ( or in the sequence of sentences in relation to a sentence in another language, the procedure for creating a database of the present invention is that each word or phrase in the first language is associated with all words and a word readings found in the selected interval of the document in the second language. This approach is also important for the reason that in one language, concepts are often expressed using more or less long phrases than in another language. The interval is determined by analyzing two documents and is used to compare the words and phrases of the second document with the words and phrases of the first document. Namely, the spacing of words or phrases in the second document is analyzed as possible associations to each word or phrase in the first document. When checking by interval in accordance with the procedure for creating a database, the number of words or phrases of the second language is determined, which can be identified and translated by words or phrases of the first language.

There are two characteristics that must be defined in order to specify in the second language document the interval in which to search for associations to any given word or phrase from the first language document. The first characteristic is the size or size of the interval in the second document, wherein said value or size is measured by the number of words in the interval. The second characteristic is the location of the interval in the second document, wherein said location is measured by the position of the midpoint of the interval. Both characteristics are user defined, however, examples of preferred embodiments are provided below. When determining the size and location of the interval, it is necessary to achieve a high probability that the interval will contain a translation into the second language of the word or phrase from the analyzed segment in the first language.

Various methods can be used to determine the size or size of the interval, including well-known statistical methods, for example, calculating a graph of the normal distribution by the number of words in a document. When using statistical methods, such as a normal distribution graph, the interval at the beginning and end of the document will be less than the interval in the middle of the document. The bell-shaped frequency graph characterizing the interval allows an acceptable extrapolation of the translation, regardless of whether it is obtained on the basis of the absolute number of words in the document or from a certain percentage of words in the document. There are other methods for calculating the interval, for example, a “stepwise” method, according to which there is an interval of one level for a certain percentage of words, a second higher level for another percentage of words and a third level equal to the first level for the last percentage of words. And again, the user can set or set all the characteristics of the interval depending on other possible parameters in order to identify useful associations to the analyzed word or phrase on

- 6 007776 first language.

The location of an interval in a document in a second language may depend on the result of comparing the number of words in the two documents. What should be qualified as a document intended for determining the location of an interval is set by the user, for example, such a document may be new articles, book chapters, and any other discretely identifiable content elements consisting of several data segments. If the number of words in two documents is approximately equal, then the location of the interval in the document in the second language will approximately coincide with the location of the analyzed word or phrase in the document in the first language. If the number of words in the two documents is not equal, then the exact location of the interval can be determined from the ratio. For example, if document A contains 50 words and document B contains 100 words, then the ratio between the two documents is 1: 2. The middle of document A is the position of word 25. If word 25 is analyzed in document A, then using this midpoint (position of word 25) as the location of the middle of the interval in document B will be an ineffective solution, since this position (position of word 25) is not the middle of the document B. Instead, the middle of the interval in document B for analyzing the word 25 in document A can be determined from the relationship between the number of words in the two documents (i.e. 25x2 / 1 = 50), manual placement in the middle of document B, or other methods.

When considering the positions of words or phrases in a document and identifying all words or phrases that fall within the above interval, the database creation procedure in accordance with the present invention provides a possible set of words or phrases in a second language document that can be translated by each word or phrase into analyzed first document. When applying the method of creating a database in accordance with the present invention, the set of words and phrases that meet the conditions of suitability as possible translations will narrow as the frequencies of associations increase. Therefore, after analyzing a pair of documents of the present invention, the frequencies of associations of words and phrases in one language with words or phrases in a second language will be formed. After several pairs of documents have been analyzed in accordance with the present invention (and therefore a large sample has been created), the procedure for creating an interlanguage associative database will produce ever higher association frequencies for any word or phrase. After obtaining a sufficiently large sample, associations with the highest frequencies give probable translations; while, of course, the limit value, upon reaching which the association frequency is believed to be accurate, is set by the user and depends on other translation methods (for example, the methods described in provisional application No. 60/276107 for US patent filed March 16, 2001 entitled Me11yub apb Arrata1i8 ίοτ Soi1ei1 Mashri1a1yui and incorporated herein by reference).

According to the above description, in accordance with the present invention, not only words are checked, but also phrases (multiple words). It was previously mentioned that phrases contain all punctuation marks and signs in the order in which they appear. After the analysis of one word in the first language is completed, in accordance with the procedure for creating the database of the present invention, an analysis of a two-word collocation is performed, then a three-word collocation, etc. with incremental building-up. This procedure makes it possible to translate words or phrases in one language, which are translated into shorter or longer phrases (or words) in another language, which often takes place. If a word or phrase appears only 1 time in all available documents in the first language, the procedure immediately proceeds to the analysis of the next word or phrase, in which the analysis cycle is performed again. The analysis stops after the analysis of all words or phrases that appear in the first language several times in all available parallel and comparable text is performed.

In a certain sense, you can combine any number of documents and process them as one document in order to find duplicate words or phrases. Essentially, so that a word or phrase does not repeat itself, it should appear only 1 time in all available parallel or comparable text. In addition, another embodiment is possible, according to which it is possible to analyze the interval corresponding to each word and phrase, regardless of whether it appears or not more than 1 time in all comparable and parallel text. In accordance with another embodiment, the database can be built by highlighting specific words and phrases that are included in the query. When words and phrases are entered for translation, in accordance with the present invention, the search for repeatedly appearing words or phrases can be performed in interlanguage documents stored in memory and not yet analyzed, by searching for interlanguage text in the network 1 and 1 through 1 using the “agents” of the search in other search engines and, ultimately, by asking the user to insert the missing association based on the results of the analysis of the query and due to the lack of a sufficient cross-language mother la.

In accordance with the foregoing, the present invention works in such a way as to analyze phrases depending on the correct placement of words (in this phrase), and can work so as to take into account the context of the selected word, as well as such grammatical features

- 7 007776 of disability, like the construction of phrases, style or abbreviations. These associations to phrases are also useful when using the double overlap translation method, which corresponds to the translation procedure described below.

It should be noted that the present invention is able to work when a subset of words or phrases belonging to a longer collocation is constantly issued as an association to a longer collocation. In accordance with the present invention, such configurations are taken into account by processing output frequencies. For example, proper names sometimes give in full (for example, “John Doe” ('EoNp Eoe)), but using the full name of the abbreviation of the first or last name (“John” (ΙοΗη) or “Dow” (Eoe)) or with another abbreviation (“Mr. Dow” (Mg. Eoe)). Since the present invention is more likely to produce single words rather than phrases (that is, it will more often give out a first or last name than the phrase “John Doe” (1oyp Eoe) of the full name), because the words that make up the phrase are inevitable If they are counted separately, and also as part of the phrase, then the mechanism of changing the rank will be applied. For example, in any document, the phrase “John Doe” (1oyp Eoe) may appear 100 times, the name “John” (1oyp) by itself or in the phrase “John Doe” (1oyp Eoe) can appear 120 times, and the last name “Dow "(Eoe) by itself or as part of the phrase" John Doe "(1st Eoe) can appear 110 times. In the normal mode of issuing a translation (in accordance with the present invention), “John” (1oyp) will have a rank higher than “Dow” (Eoe), and both of these words will have a rank higher than the phrase “John Dow” (1oyp Eoe) , in the implementation of all attempts to analyze the phrase "John Doe" (1oyp Eoe). Subtracting the number of occurrences of a longer phrase from the number of occurrences of a subset (or individual output results) allows for proper ordering (although, of course, other methods can be used to obtain a similar result). Therefore, subtracting 1 hundred (the number of occurrences of the phrase “John Doe” (1oip Eoe)) from 120 (the number of occurrences of the word “John” (1oip)) gives an accurate result for the word “John” (1oip), equal to 20. Using this analysis gives the number of appearances equal to 100 for the phrase “John Doe” (1oyp Eoe) (when analyzing and trying to translate this phrase), 20 for the word “John” (1oyp) and 10 for the phrase “Dow” (Eoe), and thus provides creating the right associations.

It should also be noted that the problem mentioned is not limited to proper names and often arises in common phrases and many different contexts. For example, each time you translate the phrase “I love you” (I 1st uey) as the most often associated phrase in another language, it is possible to independently associate the word with the word “love” (1ye) in the said other language. In addition, when the phrase is translated differently in another text to be analyzed, then again the association of the word “love” is possible (1ye). This will distort the analysis and lead to the output of the word “love” (1oue) in the second language instead of the phrase “I love you” (I 1oue uoy) in the second language as a translation of the phrase I 1oue uoy from the first language. Therefore, the system again subtracts the number of occurrences of associations to a longer collocation from the frequency of associations to all other subsets when determining the rank of associations to a longer collocation. These principles are also reflected in FIG. one.

In addition, you can put instructions in the database to ignore common words such as the personal pronoun ίΐ, the indefinite articles ap and a, the preposition о £, the adverb az, the preposition w, etc. or any common words when counting the frequencies of associations to words or phrases. This approach will provide a more accurate reflection of the true numerical values of the frequencies of associations, which in another case will be distorted due to the set of common words that appear in any given interval. This solution allows the procedure for creating an associative database of the present invention to prevent distortion of the analysis due to common words, and without unnecessary calculations of the necessary subtractions. It should be noted that if the data or any other common words are not “subtracted” from the associative database, they ultimately will not be accepted for translation anyway unless they are appropriate because they will be rejected by the double-overlap procedure, detailed the description of which is given below.

It should be noted that other calculations can be performed to adjust the frequencies of associations in order to accurately reflect the number of joint occurrences of words or phrases. For example, the introduction of an amendment excluding double counting, perhaps, would be appropriate in a situation of overlapping intervals of the analyzed words. The introduction of corrections is desirable in these cases in order to make more accurate association frequencies. The following is a typical embodiment of a method and apparatus for creating and supplementing an inter-conceptual database in accordance with the present invention based on two documents previously used as an example; and further restored the table containing the above documents.

Table 1

Document A (language A) Document B (language B) ΧΥΖΧΝνΥΖΧΖ AA BB SS AA EE GE OS SS

- 8 007776

It should be noted once again that, although in accordance with the present embodiment of the invention, the search for duplicate words and phrases is carried out in only one document, this procedure is presented only for illustration. An analysis of repeated words and phrases will be performed using the entire set of parallel and comparable texts.

When using the two above documents (document A in the first language and document B in the second language), the procedure for creating a database contains the following steps.

Stage 1

Firstly, the size and location of the interval are determined. It is indicated above that the size and location can be set by the user or approximately determined in different ways. The number of words in both documents is approximately equal (10 words in document A, 8 words in document B), so it is advisable to place the middle of the interval in a position that matches the location of a word or phrase in document A. (Note: since the ratio of the number of words in documents is 80 %, then, in another embodiment, the location of the interval can be set based on a 4/5 ratio.) In this example, the size or value of 3 for the interval can provide the best results for approximating the normal distribution graph laziness; then the interval at the beginning and end of the document will be (+/-) 1, and in the middle of the document - (+/-) 2. However, in accordance with the foregoing, the interval (or the method used to determine the interval) can only be set by the user.

Stage 2.

Next, the analysis of the first word in document A and viewing of document A are carried out in order to determine the number of occurrences of this word in the document. In the present example, the first word in document A is X: X is repeated 3 times in document A, at positions 1, 4, and 9. The position numbers of the word or phrase are simply the location of the word or phrase in the document relative to other words. Therefore, the position numbers correspond to the word numbers in the document, but excluding punctuation; for example, if a document contains 10 words, and the word “king” (cshd) appears twice, then the position numbers of the word “king” (cshd) are just the places (in a sequence of 10 words) on which the word appears.

Since the word X appears in the document more than 1 time, the procedure proceeds to the next step. If the word X appeared only 1 time, then this word would be skipped, the continuation of the procedure would be the transition to the next word and the creation procedure would continue.

Stage 3.

Possible translations into the second language are given for the first word X at position 1: the interval processing of document B gives the words at positions 1 and 2 (1 +/- 1) in document B: AA and BB (located at positions 1 and 2 of document B). All possible combinations are given out as potential translations or suitable associations to X: AA, BB and AA BB (as a combination into a phrase). Therefore, in response to X1 (the first occurrence of the word X), AA, BB and AA BB are issued as associations.

Stage 4.

An analysis is made of the next position of the word X. This word (X2) appears at position 4. Since position 4 is near the center of the document, the interval (set as described above) will be represented by 2 words on each side of position 4. Possible associations are given when considering word 4 in document B and the use of the interval (+/-) 2, i.e., the result is 2 words before the word 4, and 2 words after the word 4. Therefore, the result is the words in positions 2, 3, 4, 5 and 6. These positions correspond to the words B , CC, AA, EE and EE in document B. Taking into account all direct permutation of data words (or phrases obtained by combining them). In accordance with the foregoing, in response to X2, BB, SS, AA, EE, EE, BB SS, BB SS AA, BB SS AA EE, BB SS AA EE EE, SS AA, SS AA EE, SS AA EE EE, AA are issued EE, AA EE EE and EE EE as possible associations.

Stage 5.

Results issued in response to the first occurrence of X (at position 1) are compared with results issued in response to the second occurrence of X (at position 4), and matches are determined. It should be noted that the results that contain the same word or phrase appearing at the intersection of two intervals should be reduced to one occurrence. For example, in the present example, position 2 is the word BB; this word is issued in response to both the first appearance of X (during interval processing) and the second appearance of X (during interval processing). Since this constant position of the word is issued in response to both X1 and X2, it is believed that the word appears 1 time. However, if the same word is issued on the interval overlapping section, but from two different positions of the word, then the word is counted twice and the association frequency is recorded. In this case, the result issued in response to the word X is AA, since this word (AA) appears in both associations issued in response to X1 and X2. It should be noted that the other word that appears in both issued associations is BB; but according to the foregoing, since a given word is one and the same position (and therefore is one and the same word), achieved in the process of interval processing at the first and second occurrences of X, this word can be discarded.

- 9 007776

Stage 6.

The following position of the word X (position 9) (X3) is analyzed. The use of the interval (+/-) 1 (near the end of the document) gives associations at positions 8, 9 and 10 of document B. Since document B contains only 8 positions, the results are truncated and only the word at position 8 is output as possible values for X: CC . (Note: in an alternative embodiment, the parameters set by the user could require at least 2 characters according to the analysis conditions, which would lead to the issuance of position 8 and the next nearest position (which is the SS in position 7)).

Comparison of issued associations to X3 with issued associations to XI reveals the absence of matches and, consequently, the absence of associations.

Stage 7.

The following position of the word X is analyzed; however, the word X no longer appears in document A. From now on, the word X in language A is assigned an association frequency of one (1) to the word AA in language B.

Stage 8.

Since the word X no longer appears, the procedure involves adding 1 word, after which the phrase is checked. In this case, the phrase X Υ is analyzed, i.e. the first 2 words in document A. The method for processing this phrase is identical to the method described in steps 2-7.

Stage 9.

When viewing document A, it is obvious that the phrase X Υ appears in it only 1 time. From this moment, the adding procedure is terminated and the creation of the database does not occur. Since the end point of the procedure is reached, the analysis of the next word is performed (the specified procedure is performed whenever there are no matches for the phrase); in this case, the word in position 2 of document A is Υ.

Stage 10.

Applying the procedure consisting of steps 2-7 to process the word Υ gives the following result:

the word Υ appears twice (at positions 2 and 7), and therefore the procedure for creating the database continues (and here, if Υ appeared in document A only 1 time, then the analysis for Υ would not be performed);

the size of the interval at position 2 is (+/-) 1 word;

interval processing of document B (for position 2, which is the location of the first occurrence of the word Υ) yields positions 1, 2 and 3 in document B as a result;

corresponding words in a foreign language in the indicated issued positions are AA, BB and SS;

the use of direct permutations gives for Υ1 the following possibilities: AA, BB, SS, AA BB, AA BB SS and BB SS;

analyzes the next position of the word Υ (position 7);

the size of the interval at position 7 is (+/-) 2 words;

processing at a given interval of document B (at position 7) gives out positions 5, 6, 7 and 8, namely: EE EE SS and SS;

all permutations give the following possibilities for Υ2: EE, EE, SS, SS, EE Her, Her Her SS, Her Her SS SS, Her SS, Her SS SS and SS SS;

the selection of matching results from the set for Υ1 gives only one coincidence: SS; combining matches for Υ1 and и2 gives the SS as the frequency of associations for для.

Stage 11.

Adding at the end of the interval: since the only possible match for the word Υ (word CC) appears at the end of the interval corresponding to the first occurrence of Υ (CC appeared at position 3 of document B), the interval corresponding to the first appearance is incremented by 1, and as a result, positions 1, 2, 3, and 4: AA, BB, SS, and AA; or the following direct permutations: AA, BB, SS, AA BB, AA BB SS, AA BB SS AA, BB SS, BB SS AA and SS AA. Applying this result also gives SS as a possible translation for Υ. It should be noted that the addition to the interval is performed because the issued match was at the end of the interval corresponding to the first occurrence (the main occurrence for the word Υ); those. each time this picture occurs, in order to ensure completeness, a sub-step (or alternative step) of addition at the end of the interval will be performed.

Stage 12.

Since Υ no longer appears in document A, in accordance with the analysis procedure in document A, 1 word is added and the phrase Υ Ζ is analyzed (the next word after the word Υ is added). Adding the next word to the phrase (Υ Ζ) and repeating the procedure gives the following result: the phrase Υ Ζ appears twice in document A: at positions 2 and 7. The options for Υ Ζ at the first appearance (Υ Ζ1) are AA, BB, CC, AA VV, AA VV SS, VV SS (it should be noted that in an alternative embodiment, the interval parameters could be set in such a way as to provide for an extension of the interval size as the phrases analyzed on

- 10 007776 language A, become longer).

Opportunities for Υ Ζ in the second appearance (Υ Ζ2) are EE, EE, SS, SS, EE EE, ITS ITS SS, ITS ITS SS SS, ITS SS ITS SS SS and SS SS.

Matching gives CC as a possible association to the phrase Υ Ζ.

Expanding the interval (adding at the end of the interval) gives the following results for Υ Ζ: AA, BB, SS, AA BB, AA BB SS, AA BB SS AA, BB SS, BB SS AA and SS AA.

The application of these results also gives the SS as the association frequency for the phrase Υ Ζ. Stage 13.

Since Υ Ζ no longer appears in document A, in accordance with the analysis procedure in document A, 1 word is added and the phrase Υ Ζ X is analyzed (the next word is added after the word Ζ at position 3 in document A). Adding the next word to the phrase (Υ Ζ X) and repeating the procedure (Υ Ζ X appears twice in document A) gives the following result:

the first appearance of Υ Ζ X corresponds to the issued items 2, 3, 4 and 5;

permutations are of the form: BB, SS, AA, EE, BB SS, BB SS AA, BB SS AA EE, SS AA, SS AA EE and AA EE;

the second appearance of Υ Ζ X corresponds to the issued items 5, 6, 7 and 8;

permutations are of the form: EE, EE, SS, SS, EE EE, EE EE SS, EE EE SS SS, EE SS, EE SS SS and SS SS. comparing the two given results gives SS as the frequency of associations for the phrase Υ Ζ X; and again it should be noted that the word EE, issued as a possible association, is discarded, since in both cases it appears as one and the same word (i.e., in the same position).

Step 14

Adding the next word to the phrase (Υ Ζ X A) gives only one appearance; therefore, the creation of the database of collocations is completed and the following word is analyzed: Ζ (position 3 in document A).

Stage 15.

Using the above steps to process the word Ζ, which appears 3 times in document A, gives the following:

for Ζ1 the following results are given: AA, BB, SS, AA, EE, AA BB, AA BB SS, AA BB SS AA, AA BB SS AA EE, BB SS, BB SS AA, BB SS AA EE, SS AA, SS AA EE and AA EE;

for Ζ2 the following results are issued: EE, SS, SS, EE SS, EE SS SS and SS SS;

comparison of the results for Ζ1 and Ζ2 gives the SS as the frequency of associations for Ζ;

for the word Ζ3 (position 10) there are no results in the given interval. However, if an addition is made to the parameters that there must be at least one correspondence for each word or phrase in language A, then an SS will be issued in response to;;

comparing the results for Ζ3 with the results for Ζ1 gives the SS as the frequency of association to the word Ζ. However, this association is not counted, since the SS in position 8 of the word is already taken into account in the above association to Ζ2. If the overlap of the interval requires the procedure to count occurrences twice, the system can reduce the frequency of associations to more accurately reflect the number of true occurrences.

Stage 16.

Adding the next word to the phrase gives the phrase Ζ X, which appears twice in document A. Processing the phrase Ζ X according to the above steps gives the following:

for Ζ X1 the following results are given: BB, SS, AA, EE, EE, BB SS, BB SS AA, BB SS AA EE, BB SS AA EE EE, SS AA, SS AA EE, SS AA EE EE, AA EE, AA EE EE and EE EE;

for Ζ X2 the following results are issued: EE, SS, SS, EE SS, EE SS SS and SS SS;

comparison of the results gives an association between the phrase Ζ X and CC.

Stage 17.

Adding the next word gives the phrase Ζ X A. This phrase appears only 1 time, therefore, in document A, the analysis of the next word (X) is performed.

Stage 18.

The analysis of the word X was already performed in the first position. However, the second position of the word X, associated with another document, has not yet been analyzed for possible results for the word X. Therefore, in this case, i.e. when moving forward in a document, the word X (in the second position) is processed as when the first occurrence of the word X:

for X in position 4 the following results are given: BB, SS, AA, EE, EE, BB SS, BB SS AA, BB SS AA EE, BB SS AA EE EE, SS AA, SS AA EE, SS AA EE EE, AA HER, AA HER HER and HER HER.

for X at position 9 the following results are returned: CC.

comparing the results of position 9 with the results of position 4 gives SS as a possible match for the word X, and it is given the frequency of association.

Stage 19.

Adding the next word to the phrase (since when viewing a document in the forward direction, X appearances are no longer detected for comparison with the second occurrence of X) gives a layer

- 11 007776 sentence X A. However, this phrase does not appear more than 1 time in document A, so the procedure proceeds to the analysis of the next word (A). The word A appears in document A only 1 time, so there is a transition to a step, but not to the next phrase, because the word A appeared only once, and to the next word in document A, the word V. The word V appears in document A only 1 time, therefore, the analysis of the next word (Υ) is performed. The word Υ does not appear in any other position with a sequence number greater than position 7 in document A, so the next word (Ζ) is analyzed. The word Ζ appears again after position 8, namely at position 10.

Stage 20.

Application of the above procedure for processing the word Ζ with the second appearance gives the following results:

for Ζ in position 8 the following results are output: OS, SS and OS SS;

for Ζ at position 10 the following results are output: SS;

comparing the results in position 10 with the results in position 8 does not give associations to the word Ζ.

Again, the word SS is issued as a possible association; but since the SS represents the same position of the word, which is achieved by analyzing the word Ζ in position 8 and the word Ζ in position 10, this association is discarded.

Step 21

Adding one word gives the phrase Ζ X; but this phrase no longer appears in any (forward) position in document A, so the procedure starts anew on the next word in document A, word X. The word X does not appear more in any other (forward) position of document A , so the procedure starts again. However, the end of document A has been reached, so the analysis stops.

Step 22

The final association frequency is entered into a table combining all the results of the above procedure and subtracting the duplicate results as described.

Obviously, the amount of data is not sufficient to produce final results for words and phrases in document A. As the number of pairs of analyzed documents containing words and phrases increases along with the associations analyzed above, the frequencies of the associations will become statistically more reliable. so that related words or phrases in languages A and B will build strong associations for possible translations of words and phrases.

Program 1 (Rgodgat1), given in the attached listing of computer programs, is an example of a program designed to implement a variant of the method of creating a database. Program 1 (Rgodgat1) can be executed by a computer system of a type known in the art.

As shown, this embodiment is characteristic of the method used to create associations. The methods in accordance with the present invention are not necessarily limited only to language translation. In a broad sense, these techniques will be applicable to any two expressions of the same concept that can be associated, because, in essence, a translation into or from a foreign language exists simply as a pair association of the same concept represented by different words or collocations. In accordance with the foregoing, the present invention can be used to associate data, sounds, music, video, or any general representation that exists in the form of a concept, including concepts that can represent sensory (auditory, visual, olfactory, etc.) sensations . In accordance with the present invention, it is only necessary to carry out two implementation options (in a language translation, the implementation options are documents; in music, the implementation options may be digital representations of musical scores and sound frequencies denoting the same composition, etc.).

According to another embodiment, some rule-based algorithms well known in the art can be included in an inter-lingual association learning how to process certain classes of texts that are interchangeable in context and meaning (and sometimes can have a potentially unlimited number of derivatives), such as names, numbers and dates.

In addition, if the available interlanguage documents do not provide statistically significant results for the translation, then users can analyze possible alternatives for translations and other associations and approve and rank the corresponding selection options.

In accordance with the foregoing, the frequencies of associations between words and phrases become stricter as more and more documents in the form of translated pairs are analyzed in order to estimate the frequencies of associations. With the increase in the number of language pairs of documents to be analyzed, the method and device in accordance with the present invention will begin filling out “derived associations” between language pairs consisting of such languages that have a common association with a third language, but do not have an association directly with each other. In addition, when translated documents exist in several languages, common issued associations can be analyzed in several languages, until only one common association is found between all languages that

- 12 007776 paradise is a translation. The Rtodtat2 program shown in the computer program listing application is an example of a computer program that, when executed on a computer system of a type known in the art, represents a method in which data in the said languages are used in an embodiment of the present invention.

In addition, if the mappings in existing states are artificially assigned specific associations with particular values in a different state and the mapping data is included in the database directory, then transformations between the two states can be performed. For example, if each “concept” represented in some form, state, or in some language is assigned an association with an electromagnetic wave (tone), then an “electromagnetic representation” of the concept will be created. When a given number of concepts is encoded with the corresponding electromagnetic representations, the data (in the form of a concept) can be converted into electromagnetic waves and transmitted instantly using conventional telecommunications infrastructure. When electromagnetic waves reach the device for which they are intended, this device will synthesize individual components from electromagnetic waves and, taking into account the associations (together with the ordering instructions, using the double overlap method described in the present description, and / or other possible methods), will represent individual concepts that were presented in the form of electromagnetic waves.

Method and device for transforming concepts

Another aspect of the present invention is to provide a method and apparatus for creating a second document containing data in a second state, form or in a second language, from a first document containing data in a first state, form or in a first language, and obtain an end result consisting in the fact that the first and second documents reflect essentially the same concepts or information; wherein the method and apparatus comprise an inter-conceptual associative database. To ensure the exact conversion of concepts from one state to another state, all variants of the implementation of the translation method use the double overlap method. Conversely, known translation devices operate on the basis of the translation of individual words or use special, rule-based codes to facilitate translation from the first language into the second language. In accordance with the present invention, the use of the overlap method would allow organically linking words and phrases in a second language so that they become an accurate translation, ensuring that they have the right context and exact style in which these words and phrases are written in a second language.

According to an embodiment of the present invention, in order to provide accurate language translation, a database creation method and an overlap method are combined. Languages can be any kind of translation and are not necessarily limited to spoken and written languages. For example, the conversion may relate to computer languages, special code sets such as L8SP, etc. The database is dynamic; those. the database expands as content is introduced into the translation system, using the translation system in the process of successive iterations of previously entered information. According to a preferred embodiment of the present invention, a computing device is used, for example, a personal computer computer system of a type known in the art and widely available. However, the system does not need to use the aforementioned computing device and can be easily implemented by other means, including manually creating a database and translation methods.

The present invention can be used with a conventional computer system comprising at least a display means, an input and output method, and a processor. The display means may be any of the means widely available in the art, for example, cathode ray tube terminals, liquid crystal displays, flat panel displays, etc. The processor means may also be any of the widely available devices used in computer equipment, provided that so that the applied tool allows the computer in the process to implement the present invention. And finally, the input method used must allow the input of documents in order to build a cross-associative database; however, in accordance with the foregoing, a specific input method for converting to digital form may vary depending on the needs of the user.

Manual database creation and translation using the double overlap method

The following is a description of a typical embodiment of such a method and apparatus for translating a document from a first language into a second language in accordance with the present invention, in which a multilingual database is created by contacting the user with requests for translating words and phrases, as well as automatically generating segment translations from using the double overlap method.

In order to describe a preferred embodiment of the invention, an example is used below in which data in English is translated into data in Hebrew. These conditions are selected for descriptive purposes only and do not imply the possibility of limiting the choice of the first and second languages.

- 13 007776

In accordance with a preferred embodiment of the present invention, the computer system operates to create a database of associations between translations from English into Hebrew. The translation method comprises at least the following steps.

First, data in English is entered into a computer system.

Secondly, all entered words in English are first analyzed word by word. The database will produce famous translations of words in Hebrew. If the translation is not contained in the database, the computer system will act in such a way as to prompt the user to enter the corresponding translation. Therefore, if the Hebrew equivalent of the entered English word is not known to the database, the computer will ask the user to provide the corresponding Hebrew equivalent. Then the user will issue a translation and enter the translation into the database. For subsequent use, the computer system will work with the database in such a way that the translation is known by virtue of its input by the user at a previous point in time. Therefore, at the second stage, the input data is analyzed in a syntactically parsed state, for example, word by word, and the corresponding translations are either issued (thanks to working with the database) or entered into the database.

Thirdly, the analysis of the input data is performed in such a way as to complement the syntactically parsed segments. For example, if the data first passed word-by-word parsing, then the translation method in accordance with the present invention further analyzes the input data by evaluating two-word phrases. And again, in accordance with the above procedure, the database provides translations of two-word phrases, if known; if these translations are not known, then the computer system operates in such a way as to prompt the user to enter the appropriate translation for all possible two-word phrases. Then all overlapping two-word segments are written to the database. For example, if the phrase contains 4 words, then the database makes sure that the following combinations are stored in the database: 1,2, 2,3, and 3.4. If not, the database requests the user. It should be noted that only specially encoded translations of two-word combinations will be issued as accurate translations, even though the database will necessarily contain a definition for each word due to the implementation of the second stage described above.

Fourth, if the Hebrew translations of two overlapping two-word English phrases contain an overlapping word (or words), then the system acts in such a way as to combine overlapping segments. Redundant Hebrew segments in the overlapping region are excluded to form a meaningful translation of the English three-word phrases that are created by combining two overlapping phrases in English (and excluding redundant words in the overlapping region in English). The above steps are repeated from 1 time to an infinite number of times (n) to provide a suitable translation. The translation method works automatically by checking consistent phrases that connect coded groups of words in both languages over overlapping zones. The indicated automatically-executed approvals of the connectives, which fill the overlapping zones and are consistent in both languages, ensure the creation of a language network that is converted from one language to another language with perfect accuracy after the database reaches a critical mass.

Consider, for example, the English phrase I \ ναηΙ ίο Liu a sag ("I want to buy a car"). After processing by the method in accordance with the present invention, this phrase will be entered into a computer-controlled database. The computer will act in such a way as to determine whether the database contains Hebrew equivalents to the following words: I, \\ шИ. ίο, Liu, and sag. If these equivalents are known, then the computer will give out equivalents in Hebrew. If these equivalents are not known, then the computer will ask the user to provide the appropriate translations in Hebrew and enter the received translations into the memory for future use. Then the computer will lexically analyze the sentence into two-word segments with overlapping: I ναηΐ, \ Гап1 ίο, ,ο Liu, Liu a and a sag. The computer will act in such a way as to give equivalents in Hebrew to the given segments (i.e. equivalents in Hebrew to I \\ WI, etc.); if these equivalents in Hebrew are not known, then the computer will ask the user to provide appropriate translations in Hebrew and enter these translations into memory for future use.

Then, in accordance with the present invention, three-word segments I \\ SHI ίο, \\ SHI ίο Liu, ίο Liu a and Liu a sag will be analyzed. At this point in the procedure in accordance with the present invention, an attempt is made to combine each of the Hebrew translation pairs in which two-word English translations overlap and combine to request a translation of three-word English phrases (for example, I \\ WI and \\ WI ίο are combined with the formation of I \\ SHI ίο). If the Hebrew segments contain a common overlapping zone, which also connects them, the translation method automatically approves the conversion of a three-word English phrase into Hebrew as a translation without user intervention. If the Hebrew segments do not overlap or combine, then the user is sent a request for an accurate translation. After trying an appropriate translation of three-word English phrases, the procedure proceeds to four-word phrases, etc. and does

- 14 007776 torture by means of interlanguage overlapping automatically select combinations of translations until the analyzed segment ends (in this case, the whole phrase I \ ν; · ιηΙ ίο Liu a sag). In accordance with the method of the present invention, after performing the described lexical analysis, the translated equivalents are compared, the redundancy in the areas of overlapping segments is eliminated, and the translated phrase is displayed to the user.

Translation of documents using a database and double overlap method

In accordance with another preferred embodiment of the present invention, a document in a first language can be converted into a document in a second language using the above-described multilingual database to provide phrases for translations of words and phrases contained in the document, and then combine overlapping phrases into a second language to create a translation of the document using the aforementioned method of interlingual double overlap. For example, consider a database with access to a sufficient number of multilingual documents in order to highlight the components of the following sentence, introduced in English and intended for translation into Hebrew: Ιη abbyyup туο оеί 1ο ο отο отο д11 ί11 ίη ίονη, I a1 \ tau5 teaeb ίο е Κηονη а! с Л Л р р р № № № е е ат ат ат ат ат ат ат ат ат ат.

By performing the above procedure, the processing method can determine that the phrase Ιη abb ^^ туο that ебί Ь 1 1ίοοοόοο 1111111111111111 is the longest phrase in the source document, starting with the first word of the source document and existing in the database. This phrase is associated in the database with the Hebrew phrase Leghuha 1ίζοΐΌΗι кйей 1ΐΗϊοί ayyu a1 uebau M yaishy. The procedure will then determine the following translations using the method described above, i.e. the largest English phrase from the text to be translated (and existing in the database) with one word (or, in accordance with another option, several words), which (s) overlap (s) with the previously identified English phrase, and two translations into Hebrew for these overlapping English phrases, also with overlapping segments: 1ο \ ό6 b y a11! ys dyk ίη ίονη translates as aiu a1 uebau й baaishy bshg; 1ye dyy ίη ίονη, I a1 \ tau5 \\ 'agIeb ίο Ье 1 <ηο \ νη translates as НайайЫЫшгг, 1ннб πιίζίίν 1 ^ й ^ οί уайия; I a1 \\ ay5 \\ 'ar11eb ίο bе 1 <ηο \ νη а§ 1ее Бе-РЩРуРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРνРν of a translates as yakaykal yasY ίον й й у ра ра ра ра ί ν ν ν § § § § 1 1 пс пс пс пс пс пс пс пс <.

Given the above results issued by the database, processing will be performed in such a way as to compare overlapping words and phrases and eliminate redundant components. In accordance with this, 'Ίη abbyup ίο tu peeb ο е у еб еб а а 1 1 1 1 1 дик 1 1 ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί;;;;;; and Yueb LU a11 1ye dyke ίη ίονη is translated as ayu a1 uebau M yaayshy bshg. Using the method of the present invention, the system will receive segments in English 'Ίη abbyyug1 туο tu pss6 ίο е еб еб 11 11 11 11 ί ί и и ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ί ο ο ο ο ο ο ο ί ο ο ο ο ο ο ο ο ο ο ο ο uebau й Ь Ь ш Ы и and и у а 1 у еб еб еб Ь Ь ш Ы г г г,, and also determines the overlap.

In English, the phrases are: 'Ίη abbyyug1 ίο tu pss6 ίο е ί ί \ \ \ \ \ νη. Removing the overlap yields: 'Ίη abbyyug1 ίο tu pss6 ίο е у еб еб а 11 11 11 11!

In Hebrew, the phrases are: ег ег ^ ^ Г ί 1 ί ί ΐΌΗ ΐΌΗ ίΗί ей ίΗί ίΗί у у и и и и и и и и.. Г г г г г г г г г г г г г г г г г г г г г г.... Г.......... Г.... Removing the overlap yields: ег ю ^ ^ Г ί 1 ί ΐΌ ΐΌ ΐΌ ι й й ίΗί 1 ίΗί ίΗί ай у у а у. Г г г г г г г г г г

Then, in accordance with the present invention, the next syntactically parsed segment is processed, and thus the procedure continues. In the present example, the processing procedure is applied to the phrase 'Cheye dik ίη ίο \ νη. I a1 \ tau5 teaEb ίο Ье 1 <ηο \ νη. The system selects the English segment 'Ίη abbyug1 ίο tu pss6 ίο е е б б а а \ \!! И and a new set of English words! Corresponding Hebrew word sets are ю ю ^ Г ί ί ί ΐΌ ΐΌ ΐΌ ΐΌ г г г и и и и г и и г г и г и и и и и и и и и и и и и и и и и,,,,,,,, и,,,, и, 1, 1 и, и и и и и 1 и и 1 1 и 1 и 1 и 1 и и 1 1 1 1 1 и и и и и и 1 и и и 1 и и и и и и и и и и и и и и и. Removal of overlap in English makes two phrases: '! Ίη abbyyug1 ίο the oeeb ίο Le Ly Yueb a11 wild ίη ίονη dc and dc wild ίη ίονη, I a1 ^ au§ teaYeb ίο Le 1 <ηο \ νη in one phrase!' Ίη abbyug1 туο tu ебеб Ь еб еб 11 11 \ \ \ \ \ \ \ \ \ \ \ \ 1 § § 1 <1 <ηο \ νη.

In Hebrew, the overlap procedure works as follows. Two phrases: egyu ^ aT 1ίζοΐΌΗι vyey 1ίΗίοί ayiu a1 uebau W yaayshY shg and yaayshY shg, ίat ^ b ^ aίζ ^ ίu 1ίΗίοί uayia turn into egyu ^ aT 1ίζοΐΌΗι vyey 1ίΗίοί ayiu a1 uebau W yaayshY shg, ίat ^ a b πιίζίίν 1ίΗίοί uayia.

In accordance with the present invention, processing of the type described above continues with words and phrases remaining in the document to be translated. Therefore, in this example of a preferred embodiment of the invention, the following English phrases are 'Ίη abbyug1 1 <ηο \ νη a§! Ys bc3! p1aueg. The Hebrew translations issued by the database for the phrases cited are: ег ю ^ ί ί ί ί ίΗί г г г ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί ίΗί. Removing an overlap in English gives: 'Ίη

- 15 007776 αάάίΐίοη ΐο tu lei ΐο е е оу 1 а а 11 11 11 11 \ \ \ \ \ νπ. I a1 \ uauk \ uagIIe ΐο Ье 1 <ηο \ νπ ак Ше Бек1 р1ауег. Removing the overlap in Hebrew yields: ег ζ ζ ζ ζ ζ ζ ζ 11,,,,,,,,,,,,,,,,,,,,,,,, г г г ν ν ν ν ν ν ν ν ν ν ν ν ν ν ν ν.

By continuing the procedure following the phrase it is Ιη lyyyyup ΐο the the Levi ΐο Le 1ο \ Όά Ly a11 Ize d1g1z ίη ΐονη, I a1 \ uauk \ uagIey ΐο Le 1 <ηο \\ · η ak ek1 r1aueg Ys and Ys ek1 r1aueg ΐο e \ ^g er r1au οη 11ze Yeu Υογ1 <81a (e akkeLaP ΐeat. Relevant phrases in Hebrew are egyukaG ΙΙζοιόΟι kkek ΙίΗίοΙ lNlu a1 ueyau W kaaksh-οΐ shg (Aly SJSEC ΙίΗίοΙ uakia e1Yug kaκaΗkaη Casa ΐον and kaκaΗkaη kask1 ΐον kk kau RAAT 81kek ekνiΐζaΐ kakayigka1 кke1 тей ^ ηаΐ ие \ у \ όγ1 <. Removing the overlap in the English language gives: Ιη айй ^ ΐ ^ οη туο that zeei ΐο Ле ^ уей Лу а11 Not д1 g1k ίη ΐονη, I ауууук \ уагИе ΐο Ье Κηονη а Нен Неёк! р1ауег ΐο еуег р1ау οη ΐке № \ у Υοιί кеке уеате даете дает дает фразу фразу фразу фразу фразу фразу фразу дает фразу дает фразу фразу фразу фразу дает дает: Bshg, 1at1y ^ aΐζ ^ ίu ΗΗιοΙ wakia bе1Уг kakakkaη kask1 ΐον kk kau kaauat k1kek bekνiΐζaΐ kakayigka1 kke1 tey ^ ηаΐе \ u wig, which is a translation of the text previously proposed for translation.

At the end of this procedure, in accordance with the present invention, the translated final text is issued and output.

It should be noted that the results returned were the final processing result in a database issuing overlapping associations in accordance with the above procedure. In the course of the execution of the procedure, the system will ultimately not accept the result given in the second language, which does not contain a naturally consistent connective in overlapping with adjacent segments in the second language. If any given Hebrew result did not have an exact overlap with the adjacent associated Hebrew phrase, then this result would be rejected and replaced by such an associated Hebrew phrase that overlaps with adjacent Hebrew phrases.

The RgodgatZ program provided in the computer program listing application is an example of a program for implementing the option of manually creating a database and translating using the double overlap method. Programs can be executed on computer systems of a type known in the art.

The embodiment described above, combining the use of an interlanguage associative database with a translation method based on the principle of interlanguage double overlap, can be used in other areas to improve the quality of existing technologies by which they try to identify information in one state with information in another state, for example software known in the art of speech recognition and scanning devices with optical character recognition (OCR). Both of these technologies can verify the results obtained by the respective systems using translation methods in accordance with the present invention. If the translation does not exist and therefore an error is expected, then the user can receive a warning and a request, or the system can be programmed to search the database for such close alternatives to the non-overlapping translation that would give an overlapping translation. All the results given to the user, of course, would undergo a reverse conversion to the original language.

According to another embodiment of the present invention, there is provided a method and apparatus for creating an associative database in one language and a method and system for using an associative database to provide answers to queries or questions posed by a user. In accordance with this embodiment, the associative database can organize and store information that allows you to define and analyze associations between words or phrases. An association program may implement some of the methods of the present invention and can be used to build a database of the present invention and to analyze information stored in databases in order to determine associations between words or phrases. In FIG. 3 shows a memory 208 of a computer system 200 that stores an intelligent application 302, an association program 304, databases 306, and an operating system 308 for access from processor 202. The association program 304 may be an independent program or an integral part of the intelligent application 302. Program 304 association can analyze the database 306 to determine the associated words either in response to a request from an intelligent application 302, or in response to a request transmitted directly a user through the input device.

The system and method perform the steps of parsing the text of a document entered into the system and creating a frequency associative database in which segments of syntactically parsed text are associated with each other based on, for example, the frequency of occurrence and position of a particular fragment relative to other fragments of the document. Segments of syntactically parsed text may contain words and phrases. Documents used in accordance with the present invention can be stored in a document database for easy access, parsing and analysis of documents.

Words and phrases that often appear in close proximity to each other in

- 16 007776 document, can be used in artificial intelligence applications or with developed logic that allow the user to query the system to get an answer to a question or perform an action. The purpose of using the associative database of the present invention in smart applications is to define associations by a common third word or phrase between or among at least two words or phrases selected by the smart application.

The user can define intervals in the document database as any number of words and / or phrases near each occurrence of each selected word and / or phrase. The system then searches for words and / or phrases that are common to intervals, i.e. common third words or phrases. The frequency of occurrence of common third words or phrases in the intervals of each selected word or phrase can be stored in the frequency associative database shown in table. 3 and 4. In accordance with an alternative embodiment, the location and frequency of occurrence of words repeated in the document database can be stored in the database of repeated words and phrases, also referred to in this description as the database of repeated occurrences, an example of which is shown in Table. 5. When using the mentioned databases, the association program 304 can establish links of the highest rank for third words and phrases, in which at least 2 words or phrases are selected, selected by the intelligent application 302 based on user-defined weights or other criteria.

Building an associative database in a user-defined interval

A method is proposed for constructing an associative database of one type, referred to in the present description as a frequency associative database, which can be applied to processing documents in one language with the aim of constructing a database of related words and phrases based on their proximity to each other in the text. An example of a frequency associative database is given in table. 3. The method comprises the following steps.

a. They collect a fund of texts in one language in a “documentary database, and the larger the fund, the better.

b. Search for each word or phrase with the definition of multiple occurrences of a given word or phrase in a documentary database.

c. Set the user-defined number of words or phrases on each side of the word or phrase to be analyzed. This amount will serve as an interval. In addition to the fact that the interval is specified by a certain number of words, it can be set as wide (by all words in the specific text in which the word or phrase appears) or as narrow (by a phrase of limited size in the immediate vicinity of the analyzed word or phrase), as the user can set for specific application.

6. Determine the frequency with which each word and phrase appear in intervals near the selected analyzed word or phrase, and, if necessary, the degree of their proximity to the selected word or phrase. The following is an example of building an association between a large number of documents in one language. Proposition 1 and Proposition 2, shown in table. 2, represent two sentences among the many sentences in the documentary database and how the entire fund can be fully analyzed with all the results entered in the frequency table.

table 2

Offer 1 I MEPN TO NIOS BOSNOG APP I MAZ SPEEGBPD a Ιοί. apb not теο16 those NANA S1b apb NNE £ 1 and aye dobpd agoipb Ike saggu apb I zNoy1b gezk, keer kakhpd Ud-Naib-p S, apb a ΙϊΐΧΐβ with Unskep zoir my1bp'N NigN. Offer 2 Az and bosNog, 1't const £ ap1: 1y zeepd zpeegbpd, Geb eueb rakhepIz azkhpd "Nan NNeu sap bo But nne £ 1 and Nghe An, but I mNhsN crest, nne op1u NNhpdz NNaN gea11u mogk aga res £ APB Nhte.

The system will search for duplicate words or phrases. In these two sentences only the following words and phrases are repeated: I, ίο, ЧНе, 6oc1og, apb, δποβζίπβ, a, ώαΐ, Pi, age, gezG.

As described in the description of application No. 10/024473 for US patent, in some cases, the application in

- 17 007776 the system can be laid down instructions to consider and ignore common words of type I, a, 1o, etc. However, these common words will be considered and included in the database in cases determined by the purpose of a specific application of the system. If common words are ignored, then they will remain: Eoe1og, zpeee7shd, yi and ges1.

If it is specified that the interval contains, for example, up to 30 words on each side, the system will record the frequency of occurrence of each word and phrase within 30 words from each of these words. In addition, the system can also record the proximity of each word or phrase to the analyzed word or phrase. Since each of these words appears inside an interval of 30 words starting from each of the other words in both sentences, each word has a frequency equal to one, the joint occurrence with each of the other words, as can be seen from the table. 3, an embodiment of a frequency associative database is shown.

Table 3

Word / Phrase Associate word / phrase Frequency yosEog speegl-pd one £ 1and one geze one Zpeegtpd yosEog one £ 1and one geze one G1i zpeeg1pd one boseog one geze one KezE zpeegtpd one yosEog one £ 1and one

In addition to taking into account the frequency with which words or phrases appear anywhere within the intervals related to the analyzed words and phrases, you can build an associative database based on the frequencies of occurrence of words or phrases at a distance exactly equal to X words from the analyzed word or phrases. In these cases, the user for a specific application will need to set the exact interval in the form of 1 word or 1 phrase of a specific length at an exact distance near the analyzed word or phrase.

For example, the system can analyze the documents available in the document database to determine whether they contain a phrase up to 1 1 bf date 10,000 times, and can find a phrase up to 1 bf date within a 20-word interval relating to the word 1 e18 87 times. In addition, it can determine that a phrase up to 1o to the 1st date appeared exactly 7 words before the word 1e18 8 times (counting from the first word to the phrase).

In addition, any combination of repeating combinations of words and phrases based on the total number of words in them can be recorded. For example, in the database, you can enter the number of sentences in the database in which the word 1е1з appears at a distance of 3 words before the phrase up to the 1st to 11th date, when the word 11е1з appears at a distance of 9 words after the phrase up to the 1st to 1th date. The specified combination can appear 3 times, and the frequency of occurrence of this combination of words in the text can be used in an application that displays the meaning of the concept, either to contribute to the issuance of an answer to a question posed by the user, or to assist in fulfilling the user's request. On the basis of proposal 1 and proposal 2 in table. 2, you can create the frequency associative database shown in table. 4.

- 18 007776

Table 4

Word / Phrase Associate word / phrase Frequency Frequency of occurrence exactly through 4 words after OosTog zpeegpd one one £ 1and one 0 geze one 0 Zpeeghpd skogog one 0 £ 1and one 0 geez £ one 0 Pi zpeegtpd one 0 bosog one 0 geze one 0 Kez £ zpeegpd one 0 yosEog one 0 £ 1and one 0

As can be seen from the table. 4, from repeating words in the interval of sentence 1 and sentence 2 only the word δηββζίηΰ; repeated twice exactly at a distance of 4 words after one of the analyzed words. Data tables showing exact repeating combinations of words in the text based on their proximity to each other, measured by the number of words between them, can be created individually using a series of narrowly defined intervals.

If the above method is used to build a database of all the relationships in terms of proximity and frequency between all repeated combinations of words in the accessible text in accordance with the above description, then a large number of calculations may be required. Many of the relationships that are built as a result of this complex procedure may never be required for the application. The method described below indexes duplicate phrases to avoid pre-processing that may never be necessary.

The indexing procedure described below can be used as an alternative to the method described above for automatically determining associations by frequency and degree of proximity, and to perform general frequency-frequency analysis and analysis of exact combinations based on the location of a particular word or phrase in the interval in accordance with the above description. This embodiment of the invention is a method of constructing such a database of repeated words and phrases (or a database of repeated occurrences), which contains the location of each repeated word and phrases in a document database, the method comprising the following steps. 1. Search for repeated occurrences of all words and phrases in the accessible text. 2. Entering into the database of “locations” for each repeatedly appearing word and phrase by fixing its position in each document in which it appears, for example, by identifying the word number of the first word in the phrase and the document number in the document database. In accordance with an alternative embodiment, only the document number can be stored in the document database for the document in which the word or phrase is located. In this case, the position of a word or phrase can be determined when responding to a request. Tab. 5 is an example of information elements in a reappearance database.

- 19 007776

Table 5

Word or phrase Frequency and location ktbz Ιονθ а magt kid 20 times (word 58 / document 1678; word 4 5 / document 560; word 187 / document 45231; word 689 / document 123; ...) KShz Ιονβ xse sgeat KGZ Ιονθ and Mag Kid Kid ог д д п п Со ес! ktz Ιονε 1se sgeat Bbc khlz Ιονε sSautpd ir 1aCе bе е до х п Со Со ес! 873 times (word 765 / document 129; word 231 / document 764907; word 652 / document 4501; ...); 12 times (word 58 / document 1678; word 45 / document 560; word 187 / document 45231; ...) 10 times (word 765 / document 129; word 231 / document 764907; ...) 17 times (word 23 / document 561; word 431 / document 76431; ...) Bf £ dose. Bf bc! 684 times (word 188 / document 28; word 50 / document 560; word 7 69 / document 129; word 436 / document 76431; ...)

As shown, each occurrence of a word or phrase found at least 2 times in the document database will be added to the calculated frequency, and the location of this word or phrase will be marked with a position symbol by the word number in the document, as well as the number assigned for identification the document in which the word appears, or by using any other identifier for the location of the word or phrase in the document database.

If the database of repeated occurrences is formed entirely and completely (including numbered word positions, as well as document numbers) for all documents in the document database, then the location information allows the system to calculate any general relationships in frequency or any relationships in the frequency of occurrence of a particular combination of words, created in accordance with the above description.

In addition, if the database of reappearances has not yet been built, then the system can on the fly analyze the frequency of occurrence in at least two intervals. Any reappearance of a word or phrase that has not yet been entered into the reappearance database can be included in the database while the system responds to a request by analyzing directly the documents in the document database to supplement the analysis of the reappearance database. After the information obtained by the direct analysis of documents in the document database is used in the specific task for which this information was created, the information can be stored in the reappearance database for any future use. Whether the system performs an analysis of occurrence frequencies (frequency analysis) using a database of repeated occurrences, or the relationships are built on the fly, as a result of these actions, a database of word and word associations is created that can be used by intelligent applications.

As can be seen from FIG. 3, as part of a general frequency analysis procedure, the smart application 302 can query a frequency associative database or a reappearance database through an association program 304 by entering at least 2 words or phrases to establish which other third words or phrases often associated with some or all of the phrases presented. The system can apply two different methods when performing a common frequency analysis (CTA): either 1) an independent general frequency analysis, or 2) an associated general frequency analysis. In addition, the system can additionally perform statistics

- 20 007776 analysis after performing any of the two procedures by distributing them to an additional generation or generation of data or by combining the results and / or segments of any general frequency analysis for subsequent frequency analysis.

Independent General Frequency Analysis (1CEA)

When the smart application 302 presents at least 2 words and / or phrases for common frequency analysis (CEA) to the associating program 304, the system can identify all words and phrases often associated with the presented words using an associative database in accordance with the present invention . The system can then identify words and / or phrases that are often associated with some or all of the words or phrases represented.

The system can then use common associations between the presented words and / or phrases in a variety of user-defined ways. For example, the system can identify a common association of the highest rank by summing (or multiplying or any other user-defined weighting method) frequencies for a common association or association by a common third word or phrase to the presented words and / or phrases in a frequency associative database. Another example of a user-defined parameter that may be required is the minimum frequency (estimated by the overall rank, line number, or other criteria) for all tables of the words and / or phrases presented.

When using data elements in the example of the reappearance database, the system can determine the frequency with which 1 all sgeat and k1bz ίονβ appear in the user-defined interval in all available documents for 1 analysis, and then the system can register the frequency with which 1 all sgeat and ей э ге § §οίπ § ΐο Ьб appear together. Then the frequency of each of the independent links will be able to use the application, which will give a relative value for each link. The value will be obtained on the basis of how the frequency of the phrase 1се сgeat is estimated (according to the user's instructions in absolute or relative terms) in the frequency table of the phrase к1бз 1ονе and in the frequency table of the phrase 'эейэге § §ίίί§ί или или или, or based on the percentage of the number of occurrences association (for example, 1se sgeat) with a word or phrase (for example, к1бз Ιο \ ό) to the total number of cases of occurrence of a phrase (for example, к1бз 1ονе). Again, the frequency tables for k1bz \ο \ ό and ей ей § § § § представляют представляют представляют представляют are tabular classifications according to the rank of occurrence of third words or phrases based on their proximity to the selected phrases k1bz 1ονе and ей ΐ ΐ ΐ ΐ ΐ.

In accordance with this method, on the basis of user-defined values after analyzing the phrase 1se, the phrase can be analyzed for the relative frequency of the phrase and step by determining its position in the frequency table for k1bz 1ονе (based on user-defined requirements of the application for the width of the interval or degree of proximity), and then to determine the position of the phrase a step yid in the frequency table for the phrase ей э ге д д д.. еб. All other frequent associations (user-defined) in both frequency tables will be compared, for example, the phrase h! Outdir 1a1e, and evaluated on the basis of user-specified values of combined relative frequencies from both tables. The system will issue a phrase of the highest rank based on a user-defined weight coefficient of each frequency association. Perhaps, as a result of this analysis, the system will determine that, although k1bz Ιο \ ό (“children love”) 1se sgeat (“ice cream”) is larger than \\ agt yids (“heart hugs”), however, k1bz Ιο \ Ό ^ agt yidz (“children love hearty hugs”) is more than k1bz Ιο \ Ό 1se sgeat (“children love ice cream”) Ь э ге д ΐ (((((((((((еб ΐ ΐ ΐ ΐ ΐ ΐ еб еб (((((((((((((((((((((((((((((((((((((((((((((й больше ΐ ΐ.

As another example, when the word kanda ^ ooz and the word and phrase ίίηά and ίη Ateps are introduced into the system for an independent general frequency analysis, the table is presented. 6, which is a frequency table that can be collected from documents in a document database.

Table 6

ASSOCIATIONS AizkgaIa Kke Ζοο Jeyu 2aaaps1 capdagoosis 21000 7000 1000 ίίηά 1000 2000 500 ίη Atega; 300 5000 one hundred Total 22300 14000 1600

The word Aizyana has the highest rank, based on the sum of the total associations in rows. However, user-defined parameters can weight relative frequency values. So, by

- 21 007776 one of the possible ways, associations with the lowest frequency can be assigned a rating of one, and then associations with the highest frequencies can be evaluated depending on the multiplicity to the lowest frequency. In this case, the phrase Le / oo will have the highest rank.

Table 7

AizGgaPa Rie Ζοο Νθν 2еа1апс1 capdagoosis 21 (21 times above lowest frequency) 7 (7 times the lowest frequency) 1 (smallest frequency) mkege sap I ίίηά 2 4 one ίη Ateg1sa 3 fifty one Total 26 61 3

The relative weights of the associations show how / oo will result based on the aforementioned user-specified parameter. Similar results can be obtained by multiplying the number of times a particular word or phrase is associated with each of the queries in order to emphasize the relative equilibrium between common associations. In our example, the result is Le / oo:

1. Le / oo 7000 x 2000 x 5000 = 70,000,000,000

2. AizEaya 21000 x 1000 x 300 = 6,100,000,000

3. \ e \\ '7еа1апб 1000 х 500 х 100 = 50,000,000

To classify by rank and select associations that are common to at least 2 of the words and / or phrases presented, other user-defined criteria can be applied. This includes the assessment of some associations in some categories above others. For example, an application might assign a higher association value for a location (for example, ΐη Atepsa) to the question “where?” (For example, uneNega saη uoy GTB to ^ datok ΐη Atepsa (“where can a kangaroo be found in America?”).

Related General Frequency Analysis (CSAG)

In addition to finding common independent associations for at least 2 represented words and / or phrases, in accordance with another embodiment, a search can be performed to determine the frequent occurrences of words and / or phrases that are found in user-defined intervals, but only in documents containing at least 2 analyzed words or phrases.

This embodiment comprises the following steps.

a. The database contains the numbers of all documents common to at least 2 of the presented words and / or phrases.

b. Compare each word and phrase in user-defined intervals near all the words and / or phrases presented and record the frequency for all repeated words and phrases in the intervals. Again, a user-defined interval may be limited and may contain only repeating words or phrases in the immediate vicinity of the presented words or phrases.

For example, suppose the system is presented with 2 phrases Лбз 1ее and ЛеГоге дотд 1о Лэб for related general frequency analysis (CSGA). If the database already contains the following data elements corresponding to phrases in documents in the document database:

'k1bz 1ou a \ uagt Guide 20 times 'k1bz 1ou 1se sgeat 873 times 'k1bz 1ou a \ uagt Guide LeGoge dotd 1o Leb 12 times 'k1bz 1ou 1se sgeat LeGoge dotd 1o Leb 10 times 'k1bz 1ou 81autd ir 1a1e LeGoge dotd 1o Leb 17 times LeGoge Dotd 1o Leb 684 times

the reappearance database can direct the system to documents that are in the documentary database and contain both the phrase з з оу е е and the phrase Г Г ог от от от 1 1 еб because they have the same document number associated with them. In addition, if necessary, the system can only search for documents in which the distance between phrases does not exceed a user-specified number of words or matches any other user-defined degree of proximity

- 22 007776 to each other.

After the system has identified all documents in the documentary database that contain the phrase k1y§ 1oos in the specified proximity from the phrase bcGogs btd bc, the system can build a frequency table of all repeated words and phrases within a user-defined interval. From a limited example of the above database of words and phrases, it is known that the phrase 1cc ssat will appear with a frequency of at least 10, the phrase b1a1pd ir 1aGs will appear with a frequency of at least 17 and the phrase a teag 1shd with a frequency of at least 12. All of these frequencies can be much higher, since the corresponding phrases can appear in the same text near the phrases k1y1ouc and 'bcogs dotd 1o bc without directly adjoining them (for example, k1y§ 1ouc ccsat apy bcrc 5 \\' cc15 bcogs bdc 1o bcb).

At the request of \ ν1ιηΙ yo k1y§ 1ous bcGogs dotd 1o bc? (“What do children like before going to bed?”) You can give a lot of “right” answers. However, if there is a critical mass of documents reflecting different characteristic opinions, many higher frequencies will reflect consensus on the issue, and lower frequencies will reflect alternative opinions. For example, in the interval with the phrases k1y§1oc and bcGogs dotd 1o bc3, the phrase 1u1 kboteg may be present 3 times, which indicates the priority order of opinion, which is far from being as popular as some other opinions.

In accordance with other embodiments of the present invention, for the independent general frequency analysis (1CEA) or related general frequency analysis (JCEA), you can use a thesaurus or any other known or established phrase equivalents instead of the searched words and phrases to find duplicate words and phrases over data intervals equivalents of words. For example, the system can also search for phrases k1y§ bks, k1y§ gea11u 1oc, k1y§ cn) oo, ci1ygsp cn) oo, ci1ygsp 1oos, instead of the phrase k1y§ 1oos. The same method can be used to replace in the system the phrase bcGogs dotd 1o bc with well-known equivalents of the type of phrases bcGogs bcd, bcGogs dotd 1o Issr, bcGogs bc. A combination of words and / or general frequency methods known in the thesaurus technique in accordance with the present invention will produce a greater number of semantically equivalent words and phrases. The following is a general explanation of how the system determines semantically equivalent phrases using associative databases and intelligent application 302 and through general frequency analysis. The system can also perform independent general frequency analysis (1CEA) and associated general frequency analysis (YAEA) of the words and phrases presented and combine the results using user-defined weight parameters.

Level 2 Frequency Analysis

In accordance with another embodiment, the system can perform frequency analysis using a common association between the 1st or 2nd word or phrase and the selected 3rd word or phrase as the total frequency for the 2 analyzed words or phrases, which will add new information to the analysis results for the application. For example, if a general association, selected by the frequencies of all words and phrases within the general interval of phrases bcGogs dotd 1o bcb (1st) and k1b 1stc (2nd), is a phrase 1cc cgsat (3rd), then in accordance with this embodiment, either independent or related frequency analysis is performed for phrases or bcGogs dotd 1o bcf (1st) and 1cc csat (3rd) or k1§ 1c (2nd) and 1cc csat (3rd) and associations are selected based on the results of 2 analyzes. In addition, any at least 2 words or phrases can be analyzed using the same methods in as many combinations of as many generations of results as the user sets. Special applications will request an automated analysis to identify a common frequency analysis to be performed for each generation of association frequency analysis results. More sophisticated applications will identify at least 2 frequency analyzes that must be performed before using at least 2 independent results in combination.

Using general frequency analysis to determine the function or meaning of a word

In accordance with the description of US patent application No. 10/024473, the associations obtained between the analyzed word and phrases and the word and phrases that are found at their intervals and are not translations but are closely related concepts that can be used in artificial intelligence applications can be used with high frequency .

Artificial intelligence or an intelligent application is an application that is able to answer a question or perform a task that was not necessarily posed in the past with a requirement, respectively, to give an answer or complete a task. Associative databases can be used in accordance with the present invention to recognize word formations and / or combinations of words and phrases that can be used as class pointers to determine the function or purpose of other words or phrases associated with class pointers. For example, a particular combination of words can usually be associated with a word that is a person’s name, type of food, or action. In accordance with the present invention, using general frequency analysis, documents can be analyzed to detect the existence of these combinations and their association with other words or phrases. Based

- 23 007776 of this association, the system can classify other words that are similarly associated with these combinations, since other words that are associated in a similar way should obviously fall into similar classes. Therefore, the system can detect the presence of class pointers and use these pointers.

One example of the described application would be to assist the translation method in accordance with the present invention in making translations that the system cannot do otherwise using the above procedures. For example, suppose the system received a request in English, I 1st MozNe, for translation into X. Suppose the system has a translation of the phrase I 1stoue, but does not have translations for the phrases I 1stou Mozne or 1st mozne.

The associative database contains the word MozNe in phrases such as tupate 15 MozNe, Mg. MozNee Yesh, 1of the 18 MozNe. To identify the relationship between these phrases and the name MozNe, you can use the general frequency analysis (CEL). Then the user can indicate to the system these phrases as phrases that have a high degree of correlation with names and therefore are indicators of names, i.e. category category indicator. When the word MozNe does not appear in a user-specified number (or percentage) of phrases that are “name indicators”, the system will determine the statistical certainty that the word MozNe is potentially a name because it appears in the text, which is a “name indicator”.

Although other systems known in the art also use such name indicators and other category indicators, for example, for dates, numbers, and other special classes, the present invention makes it possible to use manually encoded indicators such as phrases reporting names to identify all other phrases in which names like MozNe and other famous names also appear. For example, the user might not have thought of adding the phrase (Not di'u paté 1z (“guy’s name”) as a name indicator, however, in accordance with the present invention, this phrase will be presented, as well as many others that the user did not turn on when using this embodiment, each word and phrase has a limited set of possible alternative words or phrases that can be found among the surrounding words and phrases. and phrases that express exactly the concepts that they replace, and will range from the original expression to thoughts that are direct opposites.For example, in the process of analyzing the phrase I 1ее с Ноео1а1е ("I love chocolate"), the system can form all possible word replacements 1ou. The system will form equivalents and almost equivalents of type I geo11u 1oue с Носо1а1е ("I really like chocolate"), I айоге с Носо1а1е ("I love chocolate"), I gea11u ep) оу с Носо1а1е ("I really enjoy chocolate"), and also nonequivalents, etc. on the contrary, type I 1o1ega (e cHoso1a1e (“I am tolerant of chocolate”) and I Na (e cHoso1ae (“I hate chocolate”). The system will not detect repeating word frequencies that do not fill everything remaining after 1ye, using an independent (or, in an alternative embodiment, related) general analysis of association frequencies. For example, databases during the analysis will not produce the phrases I ПЗН сНосо1а1е (“I catch chocolate in water”), I уо с сНосо1а1е (“I am chocolate”), I \\ 'But cНосо1а1е (“I am chocolate”).

As soon as the system determines that the word MozNe is a name, and if the phrase I 1ее is an indicator of the name (but not the only one in which the name MozNe has ever appeared), then if the translation mechanism has a translation for the word MozNe in language B, this mechanism can attach it to the translation of I 1oue in language B without overlapping. If the translation of the word MozNe in B is not known, then you can use the transliteration function from English to B to form a representation of the name MozNe in B and attach it to the translation of the phrase I 1oue into B without overlapping.

Names that are explicitly associated with other words and / or phrases as names are a simple example of how user-defined general statistical relationships between phrases that reflect concepts can be classified by their common value for use in an intelligent application. In addition, all phrases will have their own dynamic connections with all other related general concepts and concepts, which is clearly confirmed by monolingual associative databases that contain a list of related concepts based on the frequency of their being in close proximity to each other when they are expressed in a language. Each phrase that makes up a meaningful concept will have its own set of related concepts in the form of frequently repeated combinations in the text, located in a certain proximity to each other, so that the system receives the probability of values for the mentioned concepts in any given context.

If, for example, a translation request contains a word that is clearly not a name, then the frequency for related concepts can be analyzed at the next level of adjacent words and phrases to provide additional context. For example, if a request in English to be translated into language B is of the form I 1ее ЕАНН, then this request is somewhat ambiguous, since the word ЕАНН may be a name or “a sense of faith without proof”.

If other neighboring words near the phrase I 1ее ЕАНН in the full transfer request are

- 24 007776

Neg (“her”) and kNe (“she”), but not ext (“god”), hebdui (“religion”), cNigSN (“church”), etc., the system will apply formulas that will weigh the frequency , degree of affinity, as well as other category indicators in relation to associations existing in the system database, and will “understand” that one should choose the translation Faith (Rabbi) as “name” and not faith (Rabb) as “faith without proof” . The remaining neighboring words and phrases, together with their set of repeating relationships with other concepts, will provide additional confirmation of the speakers ’intentions, until the statistical significance fixes the answer or the system asks the user for clarification due to lack of information. This would happen, for example, if the whole question had the form I 1st Rabb. Since the question in this form is somewhat ambiguous, even a human translator may ask: “Do you mean that the word Rabb refers to a certain person?”

In English, the first letter of upper case implies the "name" Vera (Rabb). This property is another characteristic that the system can work with to determine that Rabb (Faith) is probably a name. Many languages that do not use the Latin alphabet do not contain upper / lower case characters, and therefore, when resolving the issue under consideration, they will only rely on level-wise relations between phrases associated with frequent proximity to each other in the text and any other representation of the language (voice, symbols, signs, etc.).

Using general frequency analysis to search for semantically similar words

Words and phrases in a language that represent a specific idea often appear in combinations. These combinations can be indicated by the frequency with which specific words and phrases are found immediately before a specific word (in English, to the left of a specific word), and also after a specific word (in English, to the right of a specific word). In accordance with this, words and phrases expressing the same concepts will have a similar type and order of words and phrases that go ahead and move away from them.

Another example of using word combinations to generate information is the creation of a comprehensive thesaurus of word equivalents and phrases. If the system is asked to identify a word that has the same or almost the same meaning as another word, i.e. the words are semantically similar, then the system can find the frequencies of words and phrases associated with the given word and search for the word in a language in which the frequencies of associations most closely match it. Usually, the more constructively similar the formation of two words and / or phrases, the more similar their meanings. Opposites will sometimes be characterized by high frequencies of common associations, but will vary greatly in some important qualitative associations that create a picture of the “signature of opposites,” which the system can also provide to applications.

The system will detect the nature of the association between any concept represented by a word or phrase, and any other concept represented by a word and phrase, the so-called “association signature”. The system uses associative databases to detect specific word formations within user-defined intervals designed to detect combinations of words surrounding a concept that defines the relationship between the concept and other concepts located in relative proximity to it. The Rgodgat4 program presented in the computer program listing application is an example of a program that uses general frequency analysis to find semantically equivalent or similar words.

For example, one embodiment using a particular word formation near a word or phrase using an independent general frequency analysis (1CPA) is to identify words and phrases that are equivalent or almost equivalent semantically (i.e., meaning ) in relation to any word or phrase. This embodiment comprises the following steps. Stage 1, which consists in the fact that a request is received for the word or phrase (query word) to be analyzed for equivalents, and a phrase (given word) of a user-defined size that appears with the greatest frequency, as well as the frequency of occurrence of the issued word is issued while the issued word is located immediately to the left of the query word in all available documents using independent general frequency analysis (1CPA). The larger the user-defined phrase, the more accurate the final results. Stage 2, namely, that an independent general frequency analysis (1CPA) is performed for each result from a user-specified number of the highest rank results of stage 1 using the interval formed by 1 word or phrase to the right of each analyzed word or phrase (using an interval of length 1 word or phrase means that the system will classify by rank the most frequently repeated words and phrases to the right of each of the words or phrases analyzed on stage 2). Then, the summation of the frequencies obtained in stage 2 of all common words and phrases is performed. Stage 3, namely, that an independent general frequency analysis (1CPA) of the request is performed using the interval formed by the user-specified number of words directly to the right of the request (and again, at least

- 25 007776 two words). Stage 4, which consists in the fact that an independent general frequency analysis (1 SEL) is performed for each word and phrase from the number of words and phrases given by the user specified by the user that were issued in step 3 using the interval formed by 1 word or phrase to the left of each analyzed word or phrases (and again, the results of applying an interval of 1 word or phrase immediately to the left of the analyzed word or phrases will be classified by rank based on frequent you words and phrases most often walking in front of every word and phrase, analyzed in step 4). Then, the frequencies of all common words and phrases obtained as a result of step 4 are summed up. In step 5, each of the words or phrases that are issued both in step 2 and in step 4 is determined. The numerical values of the frequencies of each of the common words and phrases issued in step 2 are multiplied by the numerical values of the frequencies of common words and phrases issued in stage 4. Words or phrases classified by the highest rank on the basis of the products of their frequencies issued in stage 2 and stage 4 will be words and phrases that are most semantically equivalent or similar to the query.

In the example of an embodiment with the construction of a thesaurus of equivalents of all words and phrases in the document database of the system, it is assumed that the user enters the word eRnpei (“delayed”) to determine all the equivalents of words and phrases known to the system.

At stage 1, based on the results of the highest rank (user-defined), taken in the amount of 3 for greater clarity, the system will first determine (see below) 3 phrases (user-defined), most often appearing immediately to the left of “eRnpei”:

1. 111s 5Н5рсс1 \ ν; · ΐ5 (“on suspicion”)

2. \ ν; · ΐ5 aggs51sb apb ("was arrested and")

3. sopl'shoy leb ("continued to be")

At stage 2, the system finds (see below) the words and phrases that most often follow the 3 above phrases shown as a result on the right:

1. Le 5i5res1 \ ν; · ΐ5

2. \ ν; · ΐ5 atte ^ her apb

3. by concession

a. athei (240, frequency)

b. LeI (120) with geIea8ey (90)

a. LIE (250)

b. Sopushey (150)

c. g1ea8ey (100)

a. Lea1Lu (200)

b. sopeyep1 (150)

c. Orshy (120)

Then, the frequencies of the overall results generated in step 2 are added up. In step 2, only the following general results are issued:

a. 1st 120 + 250 = 270

b. te1eagay 90 + 100 = 190

At stage 3, the system will determine the 3 most frequently occurring (according to the parameter set by the user) two-word phrases (specified by the user) immediately to the right of the selected query “eRnpei” in the documents in the database:

1. Toge diyyyuspd

2. op satdec

3. yippd le

At stage 4, the system will determine (see below) the words and phrases most often appearing immediately to the left of the 3 phrases issued at stage 3:

1. 'That diyyuspd a. 1st (300)

B. ™ ap1ey (150) with bids (100)

2. OP SATDEC a. 1st (350)

b. attei (200)

c. Lids W (150)

3. yippd Le a. LeaLiG (500)

b. 1arru (400)

c. Reor1e (250)

Then all the general results obtained at stage 4 are summarized. In this example, 2 general results are issued:

- 26 007776

a. B1b

b. Bids

300 + 350 = 650

100 + 150 = 250

Step 5. At this stage, the frequencies of the overall results issued in steps 2 and 4 are multiplied (see below):

1. b1b

2. Agg Cheb

650 x 270 = 175,500

200 x 240 = 48,000

The above illustrative example is based on a relatively small number of documents in a document database. The document database may be larger and may contain documents remotely accessible to the system over networks, for example, 1n1egpe1 networks.

The above embodiment is one of many variations of the method of the present invention that allows the use of independent general frequency analysis (1CEA) (or, alternatively, coupled general frequency analysis (KCEA)) to characterize concepts reflected by words and phrases, and the nature of the relationship between a word or phrase and any other word or phrase.

You can apply other methods based on the same fundamental idea of using the relationships that exist between concepts and are detected by their frequency and proximity to each other in the language, in combination with other category indicators, in order to solve more complex problems. These methods may include the steps of performing a general frequency analysis for at least 2 different sets of segments and using combined results in accordance with user instructions.

One example of a more complex relationship that can be detected is a combination formed by comparing signatures of words or phrases that are opposites of each other. To detect a combination, the user will enter a word into the system (for example, 1-1 (hot)). Then the system will determine all frequencies of occurrence of repeating words and phrases around a given word (as a result, a “signature” of words is formed). Then the user will turn to the system with the requirement to determine all frequencies of occurrence of repeating words and phrases around a word or phrase that reflects the opposite concept (for example, Co1b (cold)). Then, for overlapping concepts, the system will search for a common combination for the left side of the word s1b and the right side of the word s1b and for the left side of the word s1b and the right side of the word s1b1.

The result obtained is a combination formed by comparing two signatures, which the system can subsequently use to identify other pairs of words or phrases with similar combinations formed by comparisons of their signatures. Therefore, if the system receives a query in the form of a word or phrase in such a way that it searches for the opposite, then the system (1) detects all the words and phrases surrounding the query, (2) determines a list of words and phrases that have signatures similar to the query , but not with a degree of similarity that allows them to be classified as synonyms, (3) then compares the signatures of these related (but not synonymous) words and phrases with the query and (4) compares the comparison result obtained in step 3 with Signature comparisons of previously discovered pairs of opposing words or phrases. If any of the comparisons performed in step 3 contain a combination that is sufficiently similar (based on user-defined criteria) to the combination formed by comparisons of signatures of known opposites, then the system will identify a word or phrase in the results of step 2 that contradict the query so that form a combination and designate it as the opposite of the request.

It should also be noted that the parameters set by the user so that the system creates equivalent phrases (or any other relation) may contain phrases characterized by any degree of proximity to the request, and not only directly adjacent to the request from the left or right side. Correction of user-specified parameters will be required for applications in which the expression of semantic content is usually less efficient or structurally less generally accepted (for example, conversations recorded on the “forums” of the network 1p1egpe1 (sya! Goot), and other types of negotiations).

In principle, the procedure for constructing an associative database contains the steps that (ί) a unit of grouped data is organized, organized linearly or in order, (ίί) a group of data is divided into all possible adjacent subsets of the whole group, and (w) the relationships between all subsets of data based on the often repeated (usually immediate) proximity to each other of all available units of grouped data. Essentially, based on a common frequency analysis, the system provides the user with frequently repeated proximity relationships between segments of the data subsets to facilitate the detection of certain repeated combinations that define a “data signature” that provides generalized information about any data associated with the said common “data signature” . Therefore, the same procedures used to create the database, and the general frequency analysis can be applied in image recognition applications

- 27 007776 call for extracting information from data and text in various other ways. target recognition and any other application. which requires pattern recognition.

Specialists in this technical field is obvious. that if you have experience and qualifications in the above device and methods, you can make many changes. not going beyond the essence and scope of the invention.

application

ChaTsChYz '.' T 1ye '' aga ',' about £ I1e ^'dy', .e ^', Le 1 °'. 'Az'. 'Op' / Up '/ aT, Chill', '& on 'bE''ul1G,a8''on the £, dy Chez'> oi1b '' 8oi1s1 '' 5a1 (T, '1', ^i1,,, 8o _'> ab', ^'1Uo,, 'for', ^{'op1u,, eg',} a £ a 'eep', 'APB 1e''and1e!');

$ ehs1ibe_ & e = Array ( '1G' e11e '' 8Op ',' ^{Ed,,, 5e5, 'un,'} une '' E81 '' e1a11 ',' pour ',' £ 'a1ge''opA (s) heg' ₁ 'ro8eg', 'de', Che'Da'Dez. 'eG.'A'/ep'.^'.'die'/dsh'.'seY'/se'. . 'ces''NEG / seyeB' / bapz 1e ',' bapz to '/ Hodgson ^ "baa 1a', '<1 and'> gA8 Be ',' baa _'>' bargAz ^|,, ragas ',' A | '.' 1ge ',' A to au '/ aux''comme ^{1,, 51,, |} epauap!', ^'5ig,), u'>'une ^| ) keg8 ','aues')'1G5 ^, §gA (s)', 'yo1op1A (s) ^, > ^, beuo1r'_> Ά ^ ΐτε оННДА (с) ','<88ай',' (88а18 ','< I8ep! ','] E ',' ta15 ',' 81 ' _e ' oi '_>' aua1X ',' aua18 ',' aua1ep1 ',' 9sh ',' cie '_>' pop ',' zei1etep1 ',' e11e ',' eg Ιε ',' εί 1a ',' e1 cz '/ without' / bapz ');

$ ehs1ibe_zra = Array ( '1 ^°,' e11a '' ^{8u, 'un,,, sh1a'> E8>} y1e '' Zi1 ^{'horn', raga, aseg,, asep, ,,} e11o8 '' ^e11a8, 'Be', 'e1', ^{'1a,, 1o8,, |} y' ^{| a81a7ep,,} '81 ',' ^{e8e,,, que,} 'adie11o ^|,, adiePa ',' ezk '/ neo'. 'ezM .-' / egez '.' zop '.' beG. 'segsa.' a! 1abo ',' e51ag ',' zet ',' aH, 'soto' , 'epsepb1bo', 'un', 'argoBa ^, ' sop ',' bebebe ',' A (c) G, 'wo1ipid' ^, bepe ',' 11au ',' beebe ','<1uo',' yo '_; ^, rego', 'zto', 'aza -', '! ap ^, ' o ',' baa-a ', ^ u1ep', 'dshA (s) n', 'po', '8A31o' , '8o1atep1e', 1a ', ba δΐάο');

$ <Hr = LeBe;

$ <igbope = bb &ebope;

$ 1ap§ = ”.ep§;

$ o1ap £ =. Де;

$ (aye = ’bеyge;

$ 1ap§iaoesip1 = 1 $ 1apdiad = 1ap§;

$ o1apdiadesoip1 = o1ap§soip !; $ o1ap§iage = o1ap§;

# $ beSi§ = cues;

Ytsbop dekshsgoDteO {

Iz1 ($ izes, $ sec) = ehr1obe ^(And t1sgobte0); he! um ((Doa1) $ hez + (Doa!) $ hez);

} $ aPz! ag1 = dektgoyteO;

$ ίρ = о реп (((7 & 5g / 1os] / с Ь /1 /1 /1 о !, !, !, ·!

ίριιΐβ ($ ίρ, δΐ3Πίη§. бае (Н: г8). <VK> \ o);

- 28 007776 ίδίβΐίδΐ = 11е (”:р: //128.241.244.1 66 / 11з1.р11р? (Гg = $ c11g & 1ap§ = $ 1ap2); # с11ап§е $ 1etр = Ш1р1ойе (, $ 11е11з1);

$ 1151 = 8Lr1: o1o \ ueg (1pt ($ 1tetr)); $ tataggau = exp1oye (\ n, $ 11z1);

80P ($ Tataggau); ge5e1 ($ tatggau);

$ y1eaggau = aggauO;

$ sa! c = 0;

£ og ($ 1 = 0; $ 1 <coip1 ($ tataggau); $ W -) # com11 ($ tal1aggau) cbap {1G (d1e_ex1515 (81g_ger1ase ($ 1ap2, $ o1ap§, $ tatap-ay [$ 1])))) {.

$ 1eter = $ tataggau [$ 1];

$ 1tetr1 = 11e ($ tataggau [$ £ |); ipze1 ($ 1tetr2);

£ og ($ m = 0; $ m <co1 ($ 1tetr1); $ mt ++) (з £ (bt51r ($ 1tetr1 [$ m], ....)) and no1 ($ 1tetr1 [$ m]) ;

$ 1tetr1 [$ t] = ee £ 1_per1ase ([[: [time:]] +,, s1pr_1adz ($ 1tetr1 [$ t])); $ 1tetr1 [$ m] = 11йесесойе з (з з з (й__ерер11асас (& (&b.1a3; ,, $ 1tetr1 [$ m]));

ΐ £ ($ 1tetr1 [$ t]! =) $ 1tetr2. = $ 1tetr1 [$ t];

} $ d1eaggau [$ 1tetr] = and1 £ 8_epsobe ($ 1tetr2); ####### $ 1tetr = 51g_ger1ase ($ 1ap2, $ o1ap2, $ ta1paggau [$ (]); $ 1tetr1 = P1e (s1g_ger1ase ($ 1ap§, $ o1ap £, $ tatagtau [$ 1])); ! ($ 1eter2);

£ og ($ m = 0; $ m <so!! ($ 1tetr1); $ mt ++) {ί £ (s1gz1g ($ 1tetr 1 [$ t], uh! ($ 1tetr1 [$ t]);

GSG "1 = οί ^m ** s1t4g \ + aggs '/' F4-A ** m-l1 GS -" - m1 \\ ·

X [, ψΧΧ ^ YX ”/ ^ X · y) V bkAXX £ / X ['' Х Х Х Х ']] y /) $ 1teter1 [$ xn] = ig1yoseob (s1r_ger1ase (&Ыab;])); ΐ £ ($ 1tetr1 [$ t]! = $ 1tetr2. = $ 1tetr1 [$ t];

} $ Y1eaggau [$ 1tetr] = и1 £ 8_псобе ($ 1етр2);

}) (ri1z ($ f, ba1s (H: 1: z). <VU ^> bopo1o1pd i1ez! ooaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa;

$ c! oyo = sosh11 ($ ta1pa1gau);

$ ΐ = Ο;

£ og ($ 1 = 0; $ 1 <$ c! Oyo; $ 1 ++) {

ϊ £ (11e_ex151z (s1r_ger1ase ($ 1apd, $ o1apd, $ tap1aggau [$ 1]))) $ i1ex1z1 = ue; e1ze Sh1ze1 ($ y1eh1z1);

ριΐηΐ Neck = $ d1ex1z1 - $ paataggau [$ 4] \ n;

- 29 007776 ίΓ ($ ееии§ == 1гие) $ й1ех181 = ίπιβ;

ίΓ ($ β1βχΪ8ί == -! Гие) (

ίΓ ($ а 1 п аг га [[$ 1) && $ <1еи§! = ΐπιβ) {

zu81et (shu $ ta1paggau [$ 1] /i5g/1osa1/arasie/$y1yyope/.8!g_ger1ase(/i8g/1osa1/arac/e _/ $<yg/, _> $ ta1pa1tau [$ 1]));

zuyesh S'tu .z1g_ger1ase ($ 1apd, $ o1apd _e $ ta1paggau [$ 1]). / i5g / 1osa1 / arasye / $ siag <1ope / .81g_ger1ase ($ 1ap2, $ o1ap2.8! g_ger1ase ( ^, 7i8g / 1os1 / araske / $ (1n · / $ tataggau [$!])));

} $ 1n§ = $ 61 geaggau [$ tashaggau [$ 1]];

$ ο1ηβ = $ Я1еаггау [81г_гер1асе ($ 1апе, $ о1ап & $ т1п1гау [$!])];

$ 1n§8 = exp1oye (, $ 1n§);

Gog ($ ΐ = 0; $ ΐ <soip1 ($ 1pdz); $ ϊ ++) {

ϊΓ (! βΓββ ([ ^Λ α-ζΑ-Ζ], $ 1η £ 8 [$ ί])) $ 1dp [$ 1] = 8ίΓίο1ο · ννα · ($ 1η§ε [$ ΐ]);

} $ 01ηβ8 = exp1oye (, 5o1pd);

Gog ($ ϊ = 0; $ ΐ <soip! ($ О1п§з); $ ϊ ++) {

ΐΓ (! βΓβ2 ([ ^Λ Β-ζΑ-Ζ], $ ο1η§8 [$ ΐ])) $ ο1η§δ [$ ΐ] = 81t1o1ougeg ($ o1pd8 [$ 1]);

λ $ 8ite = soip1 ($ 1n £ s);

$ 8tb = soy ($ o1n§3);

ίΓ ($ 8ite> $ 8шпЬ) {Згагдт = Гойпй ($ Зите / ($ Зите - $ зшпЬ)); $ apyop - a <i;} e1ze1r ($ 5shn> $ 8ite) {$ taggsh = (goipy ($ zn / ($ znn - $ znn))); $ ayop = 8b; } eke {Zhtagdt = 1; $ asyop = ziy ”;}

ZnBn = soip1 ($ lnp3);

Gog ($) = $ 1 + 1; $] <$ s1oyo; $} ++) #tat 1ooor, goaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa ... {

1G (g1e_ex181z (81g_ger1ase ($ 1ap & $ o1ap§, $ tatggau [$]]))) # cesk

{$ d1e_81aP = ββίπύστοίπηβΟ;

ipze1 ($ apau);

Zaggau = aggauO;

$ 1ηβίρ = $ d1eaggau [$ tatggau [$ s]];

$ οΙη £ ίρ = $ i1eaggau [81g_r ^ 1ase ($ 1ap2, $ o1ap8, $ tataggau [$ 3])];

$ 1η§8ίρ = exp1oye (, $ 1pd {p);

Gog ($ ί = 0; $ ΐ <soip1 ($ 1pdz1r); $ ϊ ++) <

ΪΓ (! Едeged ([ ^l a-2A-2] ^u ₎ $ 1pd81r [$ 1])) $ 1η £ 8ίρ [$ ί] = 8k1o1o \ ¥ eg ($ 1п§з1р [$ 1]);

} $ ο1η§8ίρ = exp1oye ( ⁿ , $ o1n§f);

Gog ($ ΐ = 0; $ ΐ <sosh ($ o1n £ z1p); $ ί ++)

- 30 007776 {

ϊ £ (! ετε§ (* ( ^Λ & -ζΑ-Ζ], $ ο1ηβ8ίρ [$ ί])) $ ο1η £ 5ίρ [$ ΐ] = 8ίτίοΙολνβΓ ($ ο1η§5ΐρ [$ ΐ]);

) $ 8ΐπηβίρ = SOSH11 ($ 1pd81r);

$ 5shpk1p = soip1 ($ o1n £ 81r);

ΐ £ ($ zit1p> zitpir) {$ tagit! p = goip <1 ($ 8itepr / ($ zit1p - $ zipip)); $ acyop = ayb;} ekya ((з ит Ы <<<ите ите ф ф)))) {$ tag {pn {p = (го ип <1 ($ 5m111p / ($ 5m111p - $ 8mn1p))); $ ac11op = zb; } e1ze {$ tag§1n1p = 1; $ acnop = abb; } $ pithelf = soip1 ($ o1n§81r);

1 £ ($ 0eBid = ίηιβ) ρτίηΐ ^ eС’Н-нз'Э. ’^ VK ^;

Gog ($ ΐ = 0; $ ϊ <$ шпЪегЪ; 5ег ++) #tat Ιοορ, οονετδ ueegu zrase.

{ΐ £ ($ ί = $]) $ t = $ ί +1;

e1ze $ ηϊ = 0;

£ og ($ η = $ ηΐ; $ n <$ pitBeg1p; $ n ++) {

un8e1 ($ lzate11);

$ p = 0;

un5e1 ($ cb);

$ 1ap§81ag1 = §е1т1сойте0;

ν / bye ($ p <15 && $ 1π88 [$ ϊ + $ ρ] - $ 1η§δ1ρ [$ η + $ ρ] && $ 1п§з1р [$ n + $ p]! =) # skisk ΐ £ ϋιε $ η ινοπίδ πκίεΐι.

{$ 1ke3. = $ 1π§ ₅ [$ ΐ + $ ρ].

$ ύιεΒ1 = 1pt ($ 1bb); ϊ £ (! ετε8 ( ^,, ['~! @ # $% ^Λ & * 0 ^<> _ ⁺ = -?.,;: Λ], $ 11ιεΒ1) &&! ee ₈ ([09], 8 and 51 ($ ± еЫ, 0,1)) &&! εΓε§ ( ^Λ [0-9] * $, $ ΦεΒ1) && $ ΰιεΒ1! = && зиЗЗД ($ б1еЫ, 0,1)! = &&! еег§ ([0-9 ], 8bb51r ($ b1eB, n) && bbb1g ($ b1bb, -1)! = && bbb81r ($ bbb, 0,1)! = && 8bbb ((bbb, -1)

Ι-- mn && $ 1bfb 1 = && $ 1bbfb = &&! 1p._aggau ($ 1bbf, $ exc1is1e_ep2)) (

$ 1 meter = $ 8ΠΗγ [$ ύιεΒ1] [ΗεΒΓε \ ν_ο];

ί £ (! $ 1tetr) # ηβντ,

$ aggau [$ 1ЬеЫ] [беЬге5У_с] =}

εΙβείΓ (! 8ίΓ8ΐτ ($ ίβιηρ ,, $ ΐ,)) # ηενν, only {

$ aggau [$ ШеЫ] [118Ьг8 \ у_с ”] = $ 1 meter.

} $ ex! ha = Yaog ($ 1 / $ tag £ 1n);

- 31 007776 ΐ £ (Zasyop = h ay) {Zech! Gavt = $ ί + 5ex1ga - 45; 5ex! Ha1§ = $ ϊ + $ ex1ga + 45;

ekei (Zasyop = 5bb) {Zehgggawt - 8ί - $ ex! ha - 45; $ ex! ha! § = $ ϊ - Zekhya + 45;} ϊ £ (8ex! gawt <0) Zekhyavt = 0;

ΐ ((Zekh1ga1§> $ zt) Zekh! ga1§ = Zzit;

$ 01apy81ag1 = deitsgoyteO;

£ og ($ e = Zehavt; Ze <5ex1ga1§; $ e ++) {

Zehyap = yoog ($ n / 5gpag§t! P);

ΪΓ (Zasyop = ay) {Zo! = $ n + Zekhyap - 45; $ ίορ = $ n + Зех! Гап + 45; } e1be1 £ (Zajop = bjb) {$ bo1 = $ n - Zhejap - 45; $ Lop = $ n - ZekhSap + 45;} U (Zb0! <0) $ b01 = 0;

ίί ($ 1op> Sounds) $ 1op = Sounds;

ipve1 ($ 1bc);

£ og ($ x = 3b0!; $ X <$ 1op; $ x ++) # besk ± e bk, 10 bkb 10 Gopuagy. {ipve1 ($ 1ep§);

And (($ 1 "&& $ x> Ze) || $ 1! = $ Ϊ) # $ n> Ze &&

{$ a = 0;

Mi1e (Зо1п§в [$ е + $ а] = $ о1пдв1р [$ х + $ а] && $ о1пдв [3е + 8а] 1 =) {

$ 1ep§. =. $ o1pdv [$ e + $ a];

$ 1ep§ = 1lt ($ 1ep§);

1G (! Unified (['~! @ # $% ^A & * () о _ + = - ?. _> ;; Л], $ 1ep2) &&! Ex8 ([09], sicl1 ($! Epd, 0,1 )) &&! ezed ( ^l [0-9] * 3, $ (epd) && $ 1ep§! = && bbb1r ($! ebd, 0.1)! = &&! ezeg§ ([09], bbb1r ($ 1 bp8, -1)) && bbb1g ($ 1bb8, -1)! = && bbbb1g ($ 1bb8,0,1)! = && bbbb (31bg, -

1)! = && $ 1ep§! = && $ 1ep§! ⁱⁿ &&! t_aggau (51ep8, $ exx1iee_ & e)) {

51etraggau = aggau_keuv ($ aggau [$ 1bе1]);

ίί (t_aggau (51ep £, 81etraggau)) {

$ 1 meter = $ aggau [$ 111еЫ] [$ 1п§];

ϊ £ (! 8ίτ5ΐτ ($ ΐβιηρ, _> $ χ,)) # &&! з! г5! г ($ 1етрГ ', ₁ $ е ₎ )) {

$ aggau [$ d1e3] [$ 1п & | = $ 1tetr. $ X ,;

}}

e1we {

$ aggau [$ lbf] [$ lep8] =, $ x ,;

}

- 32 007776}

$ a ++;

} # ep <1 ο £ ν / 1ΰ1ε Ιοορ}

} # й о ог £ ог 1 1 ор ор.

} # εηά о £ ηεχν 1оор $ о! апппep (1 = DeypdsgoNteO;

$ a1 = $ o1ap§epd - $ o1ap £ 51ag1;

# phi1z ($ ph, Przeps ν / οτά ш п п $ $ п п о о 5 5 5 ег 5 1 1 р 1 ок $ $ 1 1 1)))));

} # eps1 υρ ίο 5

$ p ++;

} # εηά о £ ч? ЬЬе Ιοορ $ p <15 $ 1apdeps1 = βείτηίοΓΟίπηεΟ;

$ ite2 = $ 1ap @ ep <1 - $ 1apd81ag1;

# phi1z ($ φ, Εη§1Ϊ81ι χνοτά peter $ ΐ о £ $ peter 1оок $ 1нпе2 \ п);

}}

ΐ £ (soip1 ($ aggau)> 0) {

$ (w> 81ag1 = §е! т1сго11те0;

Hzkeat = МУ89Ь_СОМ4ЕСТ (127.0.0.1 ”,’ ’GOO1”);

$ 1eterb = argau_keu8 ($ argau) ·, ог og ($ ί = 0; $ ΐ <soip! ($ 1eter11e); $ ΐ ++) {

$ 1n§ = $ 1tepBe [$ 1];

ΐ ((8 & b3 / g_soip1 ($ a1gau [$ 1pd] [bbbcu_c] ,,) - 1> 0) {

$ 1ngc = zi3z1g_soip1 ($ argau [$ 1pd] [11ezgu_s], _> ) - 1;

$ £ etro1pd = aggau_keu8 ($ aggau [$ 1pd]);

$ η = 1;

χνΐιίΐε ($ n <soip1 ($ 1tro1pd) && soip1 ($ 1tro1p§)> 1) {

$ o1n§ = 5Itro1n§ [$ n];

$ o! ngc = 8 & b8y_soip1 ($ argau [$ 1n§] [$ o1np] ,,) -1;

$ cieuu = irda1e $ 1aEe 8βί 1o1a1 = 1o1a1 + 1, $ 1 apdiadgos = $ 1 apadiades1i + $ 1n8C, $ o1ap £ iagocess = $ o1ap8iaegesoip1 + $ o1n§c, ayc1e1 }] V) xyGee (agesk ηοί 1 £ ke '% Btashagtau ^]%' apy $ 1apgiage = '.a4Y51a8be8 ($ 1n2).' Ap4 $ o1apgiage “ ^so <M51a8be8 ($ o1pz). ');

МУ8рЬСъгаш, $ diego, $ 81geat) og <Not (# 2 Сап'1 $ cyegu .muzr_Ekv.oko);

$ pit = MU8r_AERESTEO_K.OA8 ($ 81geat);

ϊ £ ($ pit = 0) {

$ diiu - ϊπδειΐ 1§poge ΐηίο $ 1aYe ua1iez (\ 'ZH1 \, \' 1 \ ^{,,, ,,,} .aa (181a8e8 ($ 1p8) ^,,, '' aa <181a8e8 ($ o1p8).. . ^'' \ ⁱⁱ .a (1 (181a8e8 ($ 111o) .A,? '$ 1pds \ D.a0481a8e5 ($ o1pd). \, \ $ o1p8s \ \ $ tataggau [$]]

- 33 007776

KG ¥ £ RTs'gat ^', $ Diogo, $ z1geat) og with 1e ( "# $ 3 SapCh tsiegu .MU8rE_EKKOE0!);

} $ p ++;

}}

}

M ¥ 5рЬ_СЬО8Е ($ 81geat);

$ <Lbn (1 - ейшсейшссгого11шшше;

$ Hte = $ £ ϊόεηά - $ 0b3l1ag1;

phi1z ($ ίρ, άΕ 1 оook $ тее \ п);

} $ y1e_eps1 = §еПшсгоЕтеО;

}}

}}

$ apeps! = dePpGsGoNteO;

$ ite - $ a11epb - $ aPz! ag1;

£ ri1z ($ (p, 'b.

phi! s ($ φ, Epa1; ^and .a1e (V-t - <1Hl: 8). - $ ca1s - <BK> \ n); £ c1ose ($ φ);

?>

??

??

Continuation of footnote from previous page Continuation of footnote on next page

- 34 007776 <7 ίτνοπΐ = З е З З па Н па;;

$ epzra_ {= epzzra;

$ epigye_1 = ’bib;

$ Jesra_1 = Jesra;

$ c = 1;

ΜΥ8ρΕ_ € ΟΝΝΕσΓ ( ^,, 128.241.244.166, Γθοΐ);

$ Chiegu = ve1es11o (a1,1apd, 1apdsoip1, o1ap§, o1apdsoip1 iot $ epgye_1 ν / Hehe o! Ap§ = '3m'ogy';

$ gezik = M ¥ 8r (gat, 5 ^ Jehu) og Chez (Etgog # 1 - $ Jehu _W - ''.MU89_EKKOK0); $ f1egu1 = zeke! 1ap8 & ^osh $ ep§5ra_1 ν / begs o1ap§ = '$ huogy';

$ geziY1 = MU8r _(gat; $ diegu1) og <not (On Error # 2 - $ _H iegu1 - .MUZr.EKKOKO); Gog ($ ΐ = 0; $ ΐ <М ¥ 8рЬ_ЫиМ_КОА5 ($ ге8иИ1); $ ΐ ++) {

1181 ($ 1ap§) = M ¥ 8p_GETSN_K.OA ($ geziJ1);

$ ϊη. =. '. ayz ^ yesSLapd). ’;

) $ P1 = 5 & bv ($ m, 1);

$ pit = M ¥ 8рЬ_ЬШМ_К.ОА8 ($ ге8и11);

ρπηΐ $ ίη <ΒΚχΒΚ> \ η;

Gog ($ ϊ = 0; $ ΐ <$ pit; $ ί ++) {

Hz1 ($ 1o1a1, $ 1ap8, $ 1ap§s, $ o1ap8, $ o1ap§s) = M ¥ 8pE_GETSN_KOA ($ gezi11);

ρτΐηΐ ”$ 1ap§, $ Chiegu2 = ze1es1 cM yot $ jesra_1 huyege o1apd = '. aiy 1ap§ ίη ($ t) ”;

$ hezi2 = M ¥ 8pb (bgat ^, ', $ chieu2) og (e (Ergog # 3 - $ ciegu2 - .MU8 <^ E_E1 <K.OK.O); 1 £ (M ¥ 8р_ЫЦМ_КОА8 ($ ге8и112)> 0 ) {

$ ges. = $ ΐ - $ 1o1a1, Yaapd, $ 1apds, $ o1apd, $ o1a11ds <VK> \ n '';

$ c ++;

}}

ρπηΐ <VK.hVK> $ ges;

ρπηΐ $ s / ^n. М ¥ 89Ь_ЫиМ_COA8 ($ ge8i11);

?>

??

??

Continuation of footnote from previous page Continuation of footnote on next page

- 35 007776 <1

Gipsyop sopoyu ($ 5 ¥ ba1, $ o1ap§) {

ίί ($ ννή3ί = $ ™ bа1 = V;

ίί ($ 5 UYA (= \\) gate;

$ cieuu = ze1es1 z1eyeg (that. $ o1apd. 1eyeg ieege Peyeg = ’HmgLaG;

5gezik = MU5rCTmTi5i511, $ 9iegu) og <Яe (* 5> yaa1 * -nogog # 1 5cegu -. MU8р_ЕВКОК0);

; £ (MU8p_KiMKO \ Y8 (5gezik)> 0) {

Ny ($ 81eyeg) = MU8p_GETSN_K.O ^ ($ ge8iJ);

gate 5z1eyeg;

} e1ze gate 8 £ prz1a8Yez ($ 1Yaa1);

}

Y1PSYOP

0Ueg1ar ($ s, $ TT, $ teap, $ 108 _> $ o8teap, $ max, $ b1pagu_ £, $ 1ap8, $ o1apd, $ 8racea <Mge53, $ 1

$ 1tax = $ max;

5 {5o1apd) = 5ozteap;

$ {$ 1ap§} = $ theap;

$ teap = exp1obe (, $ teap);

$ ozteap = exp1obe (, $ ozteap);

£ og (5m = $ mt; 5m <soi (5theap); $ m ++) {

ipze1 (5z1pp £);

Gog ($ 1 = $ m; $ 1 <soip £ ($ teap); $ 1 ++) $ 8ίπη8. = $ Teap [$ 1].

$ 8m = $ 8 + soigy ($ teap);

ipze1 (5peh £ \ wogbta (cb);

ίί (58raceabbgesis [$ 8m + 1]) $ pehi Wogbta {cb = $ 1ap§ кеke ’.zC_her1ace (% ,, £ pt ($ s {nn §.5zraseabbgesz [5zt])). % 'og;

5ciegu1 = 8e1es1 $ 1apd, $ o1ap§ Dot $ <ysbrpagu1 chuyege $ peh1luogbta £ c $ $ 1apd =. '.8y_ger1ase (% /', d £ t (58ybpd.5zraseai (1ge88 [5zt])). 'Apb $ o1ap§ o ”ogbeg by $ 1apd meringue, 1ep§I1 (51apd), $ o1apd meringue, 1epd1y ($ o1apd) meringues;

$ gezi111 = MUZR ('' pshyBuizy ''_> ^, '$ dieguig) og b1e (sabch ergog # 2 -' 5ceguu1 '. MU8р_ЕККОК0);

$ 1 = 0;

5уЫ1е ($ 1 <МУ80Ь_МиМК.О \ ¥ 8 ($ геи111)) # && $ £ etrta1сЬ! = Уез) {

1Ϊ8ί ($ {-1shr. $ 1apv} 3 Peter ”. $ O1apv}) = MU8p_EETSN_KO ^ ($ ge5u1 £ 1); $ £ etrteap = exp1obe ( ⁿ , 5 {'Chet. 51apv});

$ 1etrotheap = exp1obe ( ⁿ , $ (1tr.5o1apd});

$ 1§ = $ t;

$ Bn = 0;

ipze1 (51strta1sy);

- 36 007776 ννΐΰΐβ ($ ΐβ <soip1 ($ 1trteal) + $ t) {

ί ((($ 8erasebbg88 [$ 8 + $ 1d] = $ 1tetrheap [$ bn]) && ($ 1tett1cb! = u)) $ 1tetre £ cb = yee;

e1ze $ 1terta1sy = by;

$ £ § ++;

$ Ip ++;

} ϊ (($ 1detter1sy = u && 8 & b1 51g_soip1 ($ 1degree ге ge8ik,) <= 8u3b1g_soip1 ($ {1tetr. $ 1apd} _> )) # cbcq ϊ £ Λβ peuyuyu ΐδ that £ cbtβ 1 1β1

{$ 1oper81ge81111 = $ {£ etr. $ 1ap§};

$ oteap = exp1obe ($ 1etrotheap [0], $ {5o1ap§});

$ 1o = soip1 ($ oteap) - 1;

Lsbesk = 8bb8y - ($ {£ etr. $ O1apd}, 0.81r1ep ($ 1troteap [0]));

ΐ £ ($ ozteap [$ 1oz] = $ 1 tetroteap [0] && 188е1 ($ o8teap [$ 1oz])) {

ΐ £ (soip1 ($ teap) + soip {($ (etrteap) -1> $ max) # δΐηβΙβννοΓά oueyar {

$ max = $ m + soip1 ($ 1trteap);

$ pz = $ t;

$ 1o1ap§ = $ {$ o1ap§}. 5bb81r ($ {1tetr. $ O1apd}, 81g1ep ($ 1tetroteap [0]));

$ oueg1ar = ίπιβ;

) e1ze $ 1terta1cb = by;

} β18βΐ £ ($ ovteap [($ 1o8-1)] = $ 1 tetroteap [0] && $ O8teap [$ 1oz] = $ 1 tetroteap [1] && 18ze1 ($ o8teap [($ 1o8-1)])) {

ΐ £ (soip1 ($ teap) + soip1 ($ 1eterteap) -1> $ max) # 3t1e \ wogb ueegcr {

$ max = $ m + soip1 ($ 1trteap); $ P5 = $ t;

$ 1o1ap§ = $ {$ o1ap§}. 5nb81r ($ {4etr. $ O1ap§), 81g1ep ($ 1etroteap [0].. $ 1troteap [1]));

$ oueg1ar = ΐπιβ;

} e18e $ 1tetra1cb = by;

} e1ze1 £ ($ o8teap [($ 1oz-2)] = $ 1 metrotheap [0] && $ ozteap [($ 1oz-1)] - $ 1 metrotheap [1] && $ ozteap [$ 1oz] = $ 1 metrotheap [2] && 188е1 ($ ozteap [($ 1о8-2)])) {ΐ £ (сшш1 ($ тап) + soip1 ($ 1етртеап) -1> $ тах) # Зтд1еугогб оуег1ар {

$ max = $ m + soip1 ($ 1trteap);

- 37 007776 $ pz = $ w;

$ 1o1ap§ = $ {$ o1ap§}. siZk ($ {'Chetr. $ o1apd}, sk1ep ($ 1 metrotheap [0]., $ 1 metrotheap [1]., $ 1 metrotheap [2]));

$ oueg1ar = 1gie;

} е1зе $ 1етрта1сЬ = по;

} e1ze1 £ (szbk_soip1 ($ {$ o1ap§},) = 0 && $ 1cbk = 1pt (ziB5 & ($ {$ o1ap§} _> 1))) {ΐ £ (coi {($ teap) + coi1 ($ 1 ) - 1> $ тах) # Зтдкчуогй оуег1ар {

$ tax = $ t + soshy ($ 1retrteap);

$ pz = $ w;

o1apdttiz zizk $ = ($ {$ 1etr o1ap _g.}, zk1ep ($ {$ o1apd}));

$ 1o1ap§ = $ {$ o1ap§}. . $ o1ap§ttis;

$ oueg1ar = cue;

}}

e1ze1 £ (siZk_soip1 ($ {$ o1ap§},) - 0 && ziZk_soip1 ($ {'Chetr. $ o1ap§},) - 0) #

{$ max = $ m + soip1 ($ 1retrteap);

$ pz - $ t;

$ k> 1ap§ = $ {$ o1ap§). . $ {'Fourth $ o1ap§};

$ oueg1ar = cue;

} e1ze {

$ 1 tetra 4cb = by;

} ΐ £ ($ oyeg1ar = kie) {

$ tteap = exp1oye (”. $ 1trteap [0], $ {$ 1ap§});

$ k> = soip1 ($ tteap) -1;

$ Po8 = sshp ($ teap) -1;

ΐ £ ($ teap [$ 1: о] && $ 1о> 0) {

$ 1sk = ziZk ($ { ^, Chet. $ 1ap§}, 5k1ep ($ 1retrteap [0]) + 1);

ί <8 Ъ Ъ с к ((1 Ь с еск еск,, 0.8 1 еп еп к (1 1 $ ((те ап ап [$ 1]]))))))))))))))))))))))))) - klp ($ те т [[$ $)))))) #

$ Iap _ё = $ {$ 1ap _ё ). 8 & b8k ($ 1cb, 8k1ep (kpp ($ tteap [$ 1o])));

}}

eKeD ($ teap [$ goats] = $ 1 tetrtheap [0] && $ teap [$ Yoz]) {

- 38 007776 $ Yap§ = $ {$ 1ap§}. si3s1r ($ {1tetr. $ 1apd}, s1g1ep ($ 1tetrteap [0]));

} e1ze {ρπηί VISION JAYAKOY; hey;}}

} $ 1 ++;

}}

ΐ £ ($ oueg! ar! = ΐηιβ) $ oueg1ar = Gake;

ΪΓ ($ 1ertax = $ max && $ oyeg1ar! = Bae) $ max = 0;

$ array = Array (s = * = * e "pin ^ - 'iiidtsfsheap) sheai of w» -> io51g_iiiptsf1i1ano,) _e o5teap => $ 1o1ap8, max => $ max, 1o1ap§ ^| -> $ 1o1ape ^{and 'oueg1ar} => $ ^oueg1ar,' '' 1op 8ez1o1apd => $ 1op £ ez1: o1ap8);

ge1it $ aggau;

} i1psyop1gap51a1e ($ \ wogs1, $ 1osk, $ 1a83, $ baz1ap8) {

§1оБа1 $ 1д_1, $ ргейх _ ^ $ ШсНопагу_1;

$ Baz1apd = bghel;

ϊί (! $ ννοτά) τβΐυτη;

ί £ ($ 4gap8ep £ = ίηιβ) {ίί (its £ ([a-gA-2], $ zuog <1)) τβίιιπι $ \ νοπ1;} ί £ ($ бае1ап§ - ееЬезеу) {$ miracle = ίπιβ; $ etaPep (1 = · oo, £ .bT ·, ϊ £ ($ ba1e1a2 = '^ arapeze) {$ sksbopagu_1 = c11syopagu] ap; $ zraez = ka1ze;} ΐί ($ ba1e1ap8 - cIbeseupp) {$ ( Psiopagusbnzip; $ serpent = ka1ze;}

M ¥ 5p_SOKNEST (216.205.78.138, ^and poO (1y) ogye (sapch soppesk. MU89_EKKOK0);

$ ινοπ1 = ίπιη ($ ννοΓά);

ΐΓ ((s1g5Y ($ y / ogf11eBgeu ($ 10_1).,) || 8ΐΤ5ΐτ ($ ννθΓ <1, Ηβ6Γβν ($ ίά._ί).) || 3υΒ8ΐΓ ($ ν / 0Γά, 3ΐΓ1εη. ($ · ΛΌΓά) - 5ίτ1εη ($ ΐά_ί)) - ЛеЬгеу ($ 1д_1)) && Ϊ8δβΐ ($ ίά_ΐ)) {

$ \ νοπ! - 5 ^ _ter1ase ([1еЬтеу ($ и_1), '\ 5' ^ from 0 ');

$ zuz1etz11e = ίηιβ;

ΐί (5ν / οτφ # \ νλν \ ν.30Γηεύιίηβ #

1Γ (8 ^ r81r ($ \ yog0,1111p: // \\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\\

$ λνοτά = 8ρΙΐΙ ('-, $ \ wogf;

$ \ νοτά = spr (&, $ \ wogs1 [1]);

hekite $ ννοΓ <1 [0];

- 39 007776 ΐί (8ίτ8ΐτ ($ · ννοΓ <1,: //)) {$ ΐβιηρ = βχρ1οάβ (: //, $ ννοπ1); 5yygezz = $ 1 meter [1]; } e1ze $ ayygezz = $ ogo <1;

BasIgezz = exp1oye (/, $ ay (1ge8z);

PayPal = $ payPal5 [0];

ί £ (! ετεβ ([3-ζΑ-Ζ .. $ aygezsbesk)) Baiygezsbesk = $ ayygesz [1];

1 £ (eeg§ ([a-gA-2], $ ac1yge83syesk)) {

$ 1ap§ = bff;

$ o1ap§ = $ base1ap§;

1x ^ & αιυιυννει \, Λαυοΐ4 ^ φνννινι. _> ν _> /)} --- ιιιι, ρ. // y χ ψγννιιχ - ии ^ ои ущ и, ι,,> ryu -}

ϊ ((Щ о ол ол у у ег (5 5 и з 1 1 г ($ $ 5 5,,, 0,8 0,8 0,8 0,8 0,8 0,8 0,8::: //:: // // // // // // // // // // // // // // // // // // // // 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 8 // // // // // // // // // // // // $ rge =

ΪΓ (8ίιΐοΙονβτ (3υΒ8ίτ ($ · ννοπ1,0,6)) = Yp7 /) {$ \ νοτά = zi3z1g ($ 1 ¥ ogy, 6); $ rge =}

} βίβεΐί ($ bае1ап§ - bеехеху) {

$ 1apd = ’bégetu;

$ o1apd = ebp;

And (siZbr ($ 5WGs1,0,8) - "§§": // *) {δτνοπΐ = siBiSr ($ x> n1,8); $ rge = bp: //; $ irreg = {gee;} e! e1 £ (ziB3l ($ 1Vy <1,0,7) - {$ · \ νοτά = ziBi1r ($ 1 Ог Og <1,7); $ pge = bp: //;} e1ze1 £ (zi3z1g ($ y / a2 <Д9) -, δδ “π: // *) {$ χνοτά = ziZi1g ($ ^ ogy, 9); $ pge = yrz: //; $ irreg = 'Chrie; } e1eI (si3i1i ($ \ woogs1,0,8) = ,, §δ “π: //) {$ \ νοτά = Ьi3i1i ($ ^ ai, 8); ргrеe = brzi: //;}.

βίββΐί (bb81g ($ lougy, 0.7) = “§“ ^ / *) {$ · ννοπ! = siZi1r ($ h ¥ op1.7); $ rge = yr: //; $ irreg ~ 1;;} e1ze1 £ (βιΛβίτφινοπΙ, Ο, ό) = {$ \ νοΓά = ЗиБз1г ($ unloading <1, b); $ pg = Dp: //;} e1ze1E (ziBz1r ($ \ yz <1,0,1) - *) {$ y / ogy = ziBzGZZ'L'ogsz); $ irreg = ΐπιβ; }} sysxxx ^ ψυ <κ> & ιαιΐ £ - {$ 1ap§ = E'arapese;

$ o1ap§ = epdk;

} ekeU ($ base1apd = cb); {

$ 1apd = cbt;

$ o1ap§ = epdk;

} $ 5 = 0;

$ 1eterger1ase = 8ίτίο1σννβΓ ($ \ νθΓά);

- 40 007776 ν / ΗϊΙβ ($ 5 <Д 1 еп еп ($! Рг рг 1 1))))) {ΐ £ (! Е е § [. [/ .. г-----]]]],,, $! Ет рг [[[[$ 8])) {

$ 1trger1ase = 8 & 3trg1ase ($ 1trger1ase, $! Etger1ase [$ 3], $ 8.1);

$ 5 = $ 8 + 2;

} $ s ++;

} $ 1eterger1ase = eger1_ger1ase ([[:: []:]] +,, $ 1erger1ase);

$ zraseabbgez8 = exp1obe (, $ 1erger1ase);

(Г- _ ЛФ »- и, Ш15е1 ($ 8rase);

$ co1og = geb;

$ soip! s = soip1 ($ 8)

$ qiegu = zeke! $ 1ap§, $ o1ap§ Dot $ b1syopagu_1 ν / yeg $ 1ap§ = '$ 1peter1ase'; $ hezik = MUZR (t1Ti8Y, $ tsiegu) og b1e (sapch heggog # 0.1 - 'Hziegue ^1. MUSSHCHYOKO);

1 £ (МУ8рДЩМкОА8 ($ ге8и11)> 0) {

From1 ($ {$ 1ap§}, $ {$ o1ap @}) = M ¥ 89L_RETSN_COA ($ gezi11);

$ vrase - $ {$ o1apd};

$ soip! s = 0;

} luyNe ($ soip! s> $ c) # wooogb

{$ 5raceabbgez8 [$ 8] =! Fri ($ 8thoseabbge88 [$ h]);

ΐ £ ($ 1a§8 = Die) {

$ open = <£ op1 co1og = \ $ co1og \>;

$ c1ose = </ £ οηΐ>;

ί £ ($ co1og - geb) $ co1og = b;

e! e $ co1og = geb;

}, Τ £ ““ A 7 \ L’P ”{

ip8e! ($ gez1oGabbgez8);

£ og ($ ΐ = $ s; $ ΐ <$ soi1z; $ ί ++) $ gez1o £ abbgez. =. Хraseеббгезз]]; $ ge8! o £ abbgezz = Dpp ($ ge8! o £ abbge88);

$ ciegu »8е1ес! $ 1ap§, $ o1ap§ Dot $ b1s ν / е $ $ д ап г = $ ез ез ез о аб б г 5 5 5 '; $ gezi11 = M ¥ 8pb (mn6i11, $ diego) og b1e (sabch ergog # 0.2 - ^ tiegu ^1. MU8P_EKKOKO);

1 £ (M ¥ 89L_KiMCOA8 ($ gezi11) = 1) {

H81 ($ {$ 1ap§}, $ {$ o1ap8}) = MU8p_EETSN_V.OA ($ ge8iJ);

$ zrase. =. $ orep. $ {$ o1app}. $ c1ose;

- 41 007776 $ race = 1pt ($ race);

$ soip15 = Ό;

} e1ze {

$ n = $ s + 1;

ipze1 ($ pehUU0gbta1cb);

ίί ($ zraseabbgesz [$ n]) $ pehk ^ ogbta1sk = $ 1ap§ nke '.51r_ger1ase (% ,, $ zraseabbgesz [$ z]). .zb_ger1ase (% ,, $ 8thеseabbgezz [$ n]).% 'og; .

$ ciegu = ze1es! $ 1ap§, $ o1ap £ Dot $ b1siopagu_1 λνΗετε $ infant Duogbta! C $ 1ap§ = '$ 5raceabbge58 [$ s]' apb $ o1ap £ o ogbegu $ 1apd meringue, 1ep§b1 ($ 1ap§), $ o1ap £ meringue , Ln £ 1b ($ o1ap§) meringue;

$ geziI = M ¥ 8pb (pnm3b, $ tzieu) og b1e (eap'1 epog # 1 - 'Ztziegu ^1. MU8P_EKKOKO);

ΪΓ ($ ta1cb = uez) ipze1 ($ ta1sk);

ϊί ((МУ8рЬЫЫМЯО5 ¥ 8 ($ ge5ik)> 0) && ($ here! = by)) {

$ n = 0;

$ tach1tit = 0;

$ tachypitt = 0;

ipze1 ($ dpa1z);

ipze1 ($ £ ta1zg);

xshze1 ($ 1op§e81o1ape); ipze1 ($ 1op§ez11ap§);

ννΗϊΙβ ($ n <ΜΥ8ΡΕ_ΝυΜΚ0ν78 ($ Γε3υ11)) # && ($ ta1sk 1 = uez)) {

Hz1 ($ {$ 1ap§}, $ {$ o1ap§}) = M ¥ 8pb_RETSN_KO \ Y ($ gezi11);

$ teap = exp1obe (, $ {$ 1ap§});

Josteap = exp1obe (, $ {$ o1ap £});

$ 1o8 = so ($ oyapeap) -1;

$ 6 = 0;

ipze1 ($ ta1s11);

χνΐύΐε ($ s <soip! ($ teap)) {

ί £ (($ zraseabbgezz [$ h $ + d] = $ TEAP [$ d]) && ($ 1 = ta1s)) = $ ta1s uev;

eke $ ta1cb = po;

$ β ++;

} ΐ £ ($ ta! cb == away) {

ίί (Д 1 еп оп ($ 1 § § 1 о ап §))) <Д 1 еп еп ($ {$ о ап §}})) $ оп § 1 1 о § § = = $ {$ 1 ап))); ίί (zD1ep ($ 1op £ e1111ap £) <zD1ep ($ {$ 1ap £})) $ 1op £ ezbap £ = $ {$ 1ap £}; Sh1ze1 (5oug1ar);

ipze1 ($ max);

- 42 007776

Zaggau = aggauO; $ agga- = ueeg! ar ($ s, 1, $ [$ 1apd], $ 1oz, $ {$ o1ap§}, $ d, $ b1scope_1, $ 1apd, $ o1ap§, 55seabbge55, $ 1op des {o1apd) ;

$ tax = 5agtau [tax];

$ 1o1ash "= $ aggau ['Чо1ап§];

ϊί ($ 1o1ap§) Sayings = ip;

νώΐβ ($ гагага [[уегегег1]]!!!! = Gake && soipDZzraseabbgezz)> Ztah) {

Zaggau = oueg1ar ($ array [8 ^n] $ array [1 Pet] 1 Pet ($ array ^[n TEAP]) $ array _[1o5]> 1t ($ array [o5tea n]) $ & 8b1siopagu_k $ 1apd, $ o1apd , $ 8raceabbge58, $ 1opde81o1apd);

ϊ £ (5aggau [oueg1ar] = ’Chgie) {

Ztah = $ aggau [tah];

$ 1o1ap§ = $ aggau [1o1apd];

Zuazok = ’Chgie;

}}

ΐί ($ max> Ztahypit && Ztah> 0) {

Ztahppit = Ztah;

$ £ pa1z = $ 1o1apd;

} ΐ £ ($ \ UAZ! = ίηιβ) {

ίΓ (81g1ep ($ {$ 1ap})> Ztahppitg) {

Ztahppitg = z1g1e11 (3 {31ap§});

$ §R = $ K $ £ ta1et = $ {$ o! Ap§};

}}

} $ p ++;

} ίί (Ззгазок · * “’ Чгие) {

$ ta1sk = ’Chgie;

And (! & 1g8 | 1 ($ & 1a18, $ 1op§e51o1ap§) && Зтппппп - - (ЗиЗзР-соиЫС, $ 1пе51о1ап2) +1)) {

# ρηηί ϊί ($ z = 0) Zrase - Zrase. $ yep. $ 1 despendo1ap§. 3c1ose;

Eke Zraze = Zraze. . Zorep. $ 1op £ eGo1ap§. $ c1ose;

$ z = $ z + zi3z1g_so_ (31opts811ap8,) + 1;

- 43 007776}

e1see {

ϊ £ ($ 8 = 0) Zrase = $ 5th race. Zorep. 5bpa1z. 3c1ose;

e1ze Zraze = Zraze. . Zorep. Zypa1z. 5c1ose;

$ 8 = $ s + ztahppit;

} ipzezztahtit);

IpzeTsZ'Uazok);

} είδβίί (Зайп1зг) ### ΐβ5ίίη§ ne \ u ώΐηβ, mgЫ Larrep8 νζΒεη and ragya1 tayek ν / az / la λ ΒΛΤνίηΐΙίΐηίΎ LA LA + 4 * "+ 1h hhr0Fa" * h \

Aiiii, α> oish ^ ishts, ish α u υνι ш шу {

ϊί ($ 8 = 0) Zrase. Zorep. $ ipa1zg. 3c1ose;

e15e Zrase = Zrase. . Zorep. Zypa1zg. 5c1ose;

Зз = $ 8 + Зет ип8е1 (3ahtpppp);

} e1ze # 8Y11-81ez £ pbase.

{ϊ £ (е е £ (([a-gA-g .. .- £], 3pracea (1c1ge88 [55])) Zrase = Zrase. Zorep. ), 5La8e1ap8). 5c1ose;

e1ze Zraze = Zraze. Zorep. cofig1 (5b3b! r (Zcraceac1c1c8e [58], 0.1) ₎ 8cbc1c8). $ c1ose;

3racease1 <1gr88 [35] = 8b3Cr (38racea <1 (1gr8z [38], 1);

1 £ (! $ 8cea (1 (1e8e [$ 3] || 3eracea <1 <1e8 [35] - 0) 3a + +;

}}

e1see {

ΐ £ (βτβ8 ([3-ζΑ-Ζ .. ZrasedeygezzzZZz])) Zrase = Zrase. . Zorep. cone1 (8 & b3e11 ($ 8sead (1ke88 [58] _> 0.1) _> 5ba8e1apg). $ c1ose;

e! ze Zrase = Zrase. Zorep. adheres to £ (8 & b3 £ r ($ 5pacac1c1cr88 [3c], 0.1), 5bc5c1cd)

BUT·

ZraSeibGezzFz] = ziBi1i ($ 8raea <1 (1r88 [$ z], 1);

1 £ (! $ Zrasead <1ge88 [58] || ЗзрасайайскеззЗЗз] = 0) Зз ++; Ipe! (8th! Cb);

}

And (’Zraseaybgesz ^ z] || 5 zraseidege88 [$ s]“ O) $ s ++;

)}

e1see {

ί £ (βτββ ([8-ζΑ-Ζ. $ zraseai <1ge88 [$ 8])) {Zrase = Zrase. . Zorep. cofee1 (8 & b3 e g ($ 8cmea <1 <1e88 [3c] _> 0.1) ₎ 5c1ose; rpp! }

- 44 007776 βίδβΐί (ετβ§ ([3-ζΑ-Ζ .. .- £], $ zraea (1 <1des [[$ s-1)])) $ drace = $ sra.. $ Orep. (55rasea <1agez8 [$ t], 0.1), $ aze1ap _g) $ s1oze.;

e1ze $ race = $ race. $ yep. copeg4 (8 & bzk ($ 5thaaaaaaaaaa [$ 8], 0,1), $ baaaaaaaa). $ c1ose;

$ 5racea1c1guzz [$ 8] = ziBzk ($ 8caeaa <kez8 [$ z], 1);

$ s ++;

ipze1 ($ ta1cb);

}} # й о £ £ ν / / Ь е Р Р ее п .. ..

$ ig1. = $ race;

ipze1 ($ 1 meter);

$ sosh ++; # ηβχΐ huogy.

ί £ (5ΪΓ5ΐΓ ($ ν / ΟΓ0, ”@)) {

$ Γβνϊά = b e ге eу ($ 1 <1_1);

ίί (zkzk ($ ig1 ,.)) $ ig1 = ege2_ger1ase (([a-gA-gO-9 / - / _ /]) @ (|> -ζΑ-Ζ0-9 / - / _ /] *) ( [/,]), \\ 1. \\ 2@tЬёбетту.сод1 \\ 3,5иг1);

eke $ ig1 = ege _ery ger1ase (([[a-gA-gO-9 / - / _ /]) @ ([η-ζΑ-Ζ0-9 / - / _ /] *) ^and , \\ 1. \\ 2@1nLeGelu.Sol1 _> $ u1);

Ipze1 ($ 8y81et8Ye);

}

1 £ (zkzk ($ \ wGl @ tEbGe \ w.s.o.)) {

$ Γβνϊά = е Ъ ге у у ($ k1_1);

$ ig1 = eee_ger1ase (([[. ..- £]). ([/ - / .. .- £ / - /] *) @ $ etayepa, \\ 1 @ \\ 2 $ geU1 (1, $ ig1) ;

} ΐ £ ($ зз1етз11е - кие &&! зк8к ($ \ уга, @)) {

ί £ ($ ep) ge1it 'bp: //. zk_ger1ase ($ ep). .n ± eLe \ CorollarySex.La1? 8uL = 8u3k ($ en <1,1);

e1ze gekip Ikr: //. 8k_ger1ase (”,, $ ig1). lpkeGelu.so.P;

} ί ((е § Ь 1 ((Ь Ь ге ^ ^ с. sol1 / ([a-2A-20-9 / -] *) / shdeh.

! 8k8k ($ duop1,1nBeGe \ u.so l1 / zat1e /)) {

$ epy = 5 & bk ($ epa, 1);

ДСзкзкфепа, ”/)) {

gyit ege§1_ger1ase ([[[: 8race:]] +,, Зк_гер1асе (-Зк_гер1асе (@, @, ($ en <1))));

}}

ge1g eeg§1_ger1ase ([[[:::]] +,, 8k_ger1ase (- zk_ger1ase (@, @, $ ig1)));

}}?>

??

??

Continuation of footnote from previous page Continuation of footnote on next page

- 45 007776> statement Βπιϊη;

>

> prog1 ₃ aua.5ts1. *;

> ίιηροΓί] aua.iH1. Weslaw,> ϊιηροΓί) aua.ni1.

> prog1 zaua.iSh.Mar;

> ππροτί) aua.i1y.Sotragaog;

> 1trog1] aua.iSh.Aggauz;

> prog1) aua.ib1.81ppdTokegi2eg;

>

> / **> * @ ийюгюгюг№№зсЬЬааа> * @ νβτδϊοη> * /> Ь Н с 1 аз аз з г г аз С С С С С С ес ап ап ап {{{>

> pnu1e ϊηί tahKeziiz;

> rpua! e 8ίπη§ hedRygase1;

> йуауа1ίίίίίίηецηецηР11ец11аз2;

> Riua1e 81ppd GedRegasEZ;

> pnu1e 81yp§ gecPbgase4;

> rpua! e ϊηί s1ag18e1Soip1;

> ryu! e ίη! epb8e! Soip1;

> ρπναΐβ ϊηί 51ag1MtCH ¥ ogb5;

> rpu1e ϊηί epbMtZUogbz;

> pnu1e boo1eap 1zSop1ep1Sgea1eb = Gake;

>

> pnu! e Veslog 51ag1pbz;

> rpua! e Ueslog epbrbs;

> ρπνβΐβ CHARACTERISTICS s1M1bryz = ηε \ ν CHARACTERISTICS (10);

> rPaGe nazyEe eMlbryz = ne \ y NazyEye (U);

> пу 1 Наз Наз Ы Наз = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = = =;

> пупу11 Наз Наз Наз Наз Наз Наз Наз Наз и и и и и и и иРЫРЫРЫ = = = = = = = pet / Naz / /Ь петее 10 (10);

>

>

> пу 1 Соп Соп Соп п п п б оп оп оп оп оп п = = = pi11;

>

> / ** ге а ез ез η аз аз аз С С С С С С>>>>> * * * * * ап 1 1 б б б 1 1 1 1 1!! bailee> * /> billet bruceceo11eclBea0 {> maxxCeNe3 = 30;

> gcdbgase1 =

- 46 007776> hechRygase2 => hechRygase3> hetRygase4 => 81ag18e1 Sochi = 4;

> epy8e1Cosh11 = 4;

> 51ag1M1nAogY5 = 1;

> epyMtAogyz = 1;

> bSoppesyop = Vash.deyUVSoppesNopO; >)>

> Ь с ί ί ΐ ΐ ΐ 1 М М К К К з з θ>>> ге т т ах ез ез и;;;

>}>

> bilobus 8thisnum 8echschrrb8e1 0 {> geelite hetrbr8e1;

>}>

> ЪННΐΐΐΐηη § §ее11ечечечРЬЬ888ее 0 0 0 {{{{{>>>> ге ге!! гедедедедедедедедР22222;

>}>

> RiNs 8ίπη§ eeSheyaRgazeZ 0 {> gs1it gedRgazeZ;

>}>

> Ь И ΐ ΐ ΐ § е § § е 1 д 4 Р 4 4 4 4 0 0 {4;

>}>

> Ы г с ί ΐ ΐ 8е181ΐ1818еСошеош О О О О {{

>}>

> ЪННί ίηΐ §е1Epy8e1Sosh110 {> gyit edy8e1Soip1;

>}>

> Ъηΐ ΐηΐ §е181айМтАогйз О {> г1ит зкаГЛтАогйз;

>}>

> RiNs ίηΐ βεΐΕπάΜίπΑοτάδ 0 {> ge1it epyMtAogyv;

>}>

- 47 007776> ЪННίίίίηη []]] ]ссс11РРРРазООО {>>>>> >ίίίίίРί§§]]]]] ап ап = ~ ~ ~ ~ ~ ~ ~ гецРецРгагага55е,, гедедРЬгаа222, гея гедЪггЗЗ,, г гедедРк444}}; > ge! igp up;

>}>

> pbNs νοίά ССССахККезиППП (ΐηί x) {> max.

>}>

> pbnc νοίά 5β £ ςΡ.βςΡ1ΐΓ88β1 (8ίτίη§ x) {> heptBlase1 = x;

> //8u81et.oSh.rppip (No net-1-btd! ..);

>)>

> Ъ Н ίά ίά!!! K. ея Р Ь г 2 2 2 (х ί г ® ® х x) {> г ея Р аз аз 2 2 = x;

> //8uz1et.oi1.rpp11n(Hetu wa1ie-P bld βει ..);

>}>

> ЪНН νν!!!!! !ецецРггазазЗЗЗ 81 81г 81гт 81 81 81 81 81 81 81 81 81г 81 х 81 х 81 81 81 81 81 81 81 81 81>>> >едедедедгагааз = = = x;

> //8ua1et.oi1.rhtnn («« «« «у § § § § § § .. .. .. .. ..);

>}>

> Ъ Н с ν ίά ίά с с с В. ц ц РИ РИ 4 4 4 ((З З п д д д х {{>> ед ед Р аз аз 4 4 4 4 4 = x;

> //8ua1et.oi1.rgtnn (е у у у 1--ет д д!! ..);

>}>

> ЪННс νίάίάίά!! !аг !агаг! 8ссссСошош1111 (ΐ (ΐηίί) {>!! аНН888оипип11 =;;

>}>

> ЬЬНс νοίά βεΐΕηάδεΙΟοιιηΐ (ίηί x) {> εηάδείϋουηί = x;

>}>

> ������ νοίάεείδίβΓίΜίη'νν'ΟΓάΒ (ΐηί х) {> с ШМтюТяйз = x;

>}>

> ЬЬНс νοίά ββΐΕηάΜίηνΤοιάβ (ϊηΐ x) {> cnnmt ^ Wogbz = x;

>}>

> pnu1e ΐηί §е1 РиггееИ (81лпд рБг) {

- 48 007776> 1r (pbr = pi11) pbr =;

> 1 ^ {> 1 £ (! (Pkg =)) {>

>} e1see gest 0;

>} ca1cb (Exseriop e) {> 5c1e1e1.o1.ppp11n (Exci11ci eGePeSeffe Sérgase GO & from> Brg.GeGe (PgaseCo11ec1Beap.Ge1PeGe1c1): + e);

> г з = Е = сер сер = Н Н де де де де е ГО ГО ГО ГО ГО ГО ГО & от от>> В ат ат ((((((Р Р г е С С С ап ап ап ап))))))):: + e; > ge1it 0;

>}>}>

> пу 1 Мар Мар Мар.... [[]]] 80 80 80 80 Наз Наз Наз Наз Наз Наз Наз Наз Наз Наз Наз Наз Наз Наз Наз>>>>>>> уа уа а .8 .8 .8 .8 .8 1 = = = = = = =.

> Mar. Ep1gu [] egipes = (Mar. Ep1gu []) e1.1oAggau (ne \ y> Mar. Ep1gu [5e1.81geO]);

> Atgauz.zogfepyes, ne \ y Sotragaog () {> pbnc ϊηΐ on the side of the (O]] es1 o1, O]] es1 o2) {> O1ybc1 ob) 1 = ((Mar. Epnu) o1) .2e1Ua1ieO;

> О1] es1 o1c2 = ((Mar. Ep1yu) o2) .§е1Уа1еО;

> he! um ((Sotraga1e) b] 2).

>}>});

> hellit

>}>

> Ь И с ΐ ΐ ί ί ос ς з 1 1 1 {ос ос 1 1 {{{>>> о § з з>>> 18 ге ге ге 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18

> K. ezi118e1 p2K. Ezi11 = pi11, p1B. Ezi11 = pi11;

> S1M1c1Pb8. c1eagO;

> epMk! Rkgz.s1eagO;

> itz81Pbr8. s1eagO;

> ityaEpRgz.skagO;

> 81pp§Vi £ Sez gezi1181g = ηβιν δίπηβΒιιίϊβΓζ);

> δΐτίηβ 1 Я Я Я е и 11 = 11 = = => Г аг аг 1 Ь г г = = pemg г 1 1 ог ог ((аг 18 18 е ош ош));

> epBr3 = oven Ueslog (enes18e1Coip1);

>

> Sopesyop bSoppesyop = pi11;

>

> ΐηί ρϊά = 0;

> δΐτϊηβΟ pgasez = aeFbasezO;

- 49 007776> 8ΐιϊη§ srgu => ίοτ (ΐη1ΐ = 0; · ί); .

> ge8i1131g.arrepb (+ ϊ +);

> gezi1181 g.reprep (+ ρίσηΐ +);

> 8ίπη§ Ιβπιρδί = р1Рг;

> ϊ £ ζ (δί = s1ag (MtAog <1z)) {> ιηΐ ρΐΐά == § е Р аз аз аз 1 1 1 (p1Pr);

> ϊί (ρ1ϊά! = 0) {> 81agL11gz.absSh1etep1 (+ p11s1);

> 8ί ++;

> Gezi81g.arrepb (+ ρ1Ρτ> +);

>}>} e1ze {> gesi1181g. arrep (1 (+ p1Pr +);

>}> heziZ & .arrepbS + (81pp8) s1M1bR11G8.de1 (p1Pg) +> heziK8 (g.arresh1 ("”);

> Gezi1181g. arrebp (+ (8 £ rt§) enM1 <1Pbzi§e1 (p2Pr) +> ”);

>

> 8ίπη§ 1eterEn = p2Pr;

> 1 £ ((en = enbMtAogbz)) {> ίηί ρ2ϊά = 8e1Pbase1 <1 (p2Pr);

> ϊί (ρ2ϊά! = 0) {> б б г Ь г з, and <1bE1etep1 (+ p21b);

> ep ++;

> Gez1Y £ Arrepb (+ p2Pr> +);

> r 'l>} e1ze {> hezik81g.arrepb (+ p2Pr +);

>}> hezik8b-.arrepb (* + p2sp1 +);

> gezi1181g.arrepb (p);

> ί ++;

>}> t ++;

>}>

> ϊί <ϊ> 1) {> sgea1eAzskhdpe (1WO;

> sta1eTsa1aW (s1e1n1e8, eniLe, zlibrz, elMkIiz);

> 1sCop (en1Crea1eb = Ga1ze;

>}

- 50 007776> / ** />} ca1cb (Exseriop e) {> geziu81g.arrep <1 (Exseriopi ΐ ΐ ίΐ>>> ((г аз е С 11 ес ес ес ап) ((Ле 1 1 + 1 + + ;

> //8uz1esh.oi1.rgJn (Exeriop λνΐύΐβ ге пе ут ут § § г г>>>>> (RygaseSo11es1Beap.de1Az80sVu1o1P8) + e);

> оггогогз + + = сссербербоп у у ЛП Γβΐπβνίηβ г Ле г г>>>>> (RigaseCo11ec1Beap.de1Az5ocBu1occ) + e;

> e.rpp181askTgase ();

>}> ge1it gesiISDAAOZLtdO ί> ί>

> Ь Н ί ί § η η § де 1 К т 1 ез ез ез ((1 1 ах З оч у у {) {> г ог з = => 1 § § з § § =;

> 81lp§ 5ΐΓ1 = 3ΐτ2 = zDZ => Rgerageb81a1etep1 p1Ke18L> 1 = pi11;

> ίηΐ peu / Ssoge = 0;

> REAL 1pcM1bRiggs = ne \ ¥ NazYaEeO;

> CHAIN RbPbz ^β peu / CHARACTER;

> NazKePb21 <18 »πβνν NazKaEeO;

> OpenGuideGe1Kezi118D = ηβιν> OpenGuideS ’Sottop Rgasez т zihgopbshd ge1abue1u sottop ιηίάάΐβ Ьgase> 8. ΝΟΤοΛΙ ЗсогерГгазе 2);

> ίηί evil = 1;

>

> 1gu {> 1G (bSoppes11op "" "piP) {> //8uz1et.oi1.rppL1(’Т>В Согт Ϊ8 Νιιΐΐ ϊη де (К.е1аУюез! '');

> bSoppesyop = Vgat.deyuVSoppesyopO;

>}>} sa1cb (Exseriop e) {> //8uz1et.oi.rppL1 (Exser1yup y / Vdedeysd sopeseop Dot> Vgat: + e);

>}>

> u ((s1ag1Pbg8 = pi11) || (e1ag1P11G8.8YeO = 0) || (ebpBbz -> pi11) || (βηάΡΗτβ.είζβΟ = 0))

> 8ΐπη§ 51ag1Sopb => Gog (ϊηί ΐ = 0; ΐ + Зпо +);

> 8ΐπη§ Й§ЦЦЦ = ηβνπν =η + т;

> ge1Be8i118D.arreps1 ('H rbgZsoge + + rrgKeu +);

> zpo-n-;

>}> // 8v1 en!! And {.ρπιιΐ1η (”Ι8 СпГепГ Сеа1еb ??: +

- 51 007776> 15Sop! Ep! Sgea! Eb);

> ΐί ((! 1cSoop! eGeSgeakeb) && (тахЗЬсж! = 50) && (evil> 1)) {> srea! eSop1ep1vt (en! ne8);

>}>} sa1cb (Exser! 1op e) {> //8uz!et.oi.rpp11n (Exseriop y / b1e exesialpiguet: + e);

>)> ge1it ge1Kesi118y.1o81ppёO ⁺ >}>

> ρπνβίβ νοίά сеа1еАз81§пебВ0 {> ϊί (! азз1§бeнзлзETр1уО) аз51§пе <1На8Ь.с1еаг0;

> a85A8pebNaa! 1. p \ 1!

> £ οτ (ΐηί ί = 0; ϊ

Claims (9)

CLAIM 1. A method for associating words in a language, comprising the steps of creating a fund of documents, said fund comprising at least one document; choose the first word or phrase and the second word or phrase; find many occurrences of the first word or phrase and the second word or phrase in said fund; specify in the said fund the first intervals and second intervals, while the first intervals contain the first word or phrase, and the second intervals contain the second word or phrase; searching for said first intervals and second intervals for common words or phrases, wherein said common words or phrases appear in a plurality of intervals; and associate the first word or phrase and the second word or phrase with common words or phrases based on the frequency of occurrence of common words or phrases within, respectively, the first intervals and second intervals. 2. The method according to claim 1, in which the said association of the first word or phrase and the second word or phrase is improved by a higher frequency of occurrence of common words or phrases. 3. The method according to claim 1, wherein said association of the first word or phrase and the second word or phrase is improved by a lower frequency of occurrence of common words or phrases. 4. The method according to claim 1, further comprising the step that the first word and / or phrase is replaced with a substantially semantically equivalent word or phrase. 5. A method of associating words in a language, comprising the steps of creating a fund of documents, said fund comprising at least one document; choose the first word or phrase and the second word or phrase; find all documents containing a plurality of occurrences of the first word or phrase within a predetermined proximity interval of the second word and / or phrase, wherein said predetermined proximity interval is bounded by an upper limit and a lower limit; set the interval in the documents found, while this interval contains the first word or phrase and the second word or phrase; search for said intervals for common words or phrases; and associating the first word or phrase and the second word or phrase with common words or phrases based on the frequency of occurrence of common words or phrases within the above-mentioned intervals. 6. The method according to claim 5, in which said association of the first word or phrase and the second word or phrase is improved by a higher frequency of occurrence of common words or phrases. 7. The method according to claim 5, in which said association of the first word or phrase and the second word or phrase is improved by lowering the frequency of occurrence of common words or phrases. 8. The method of claim 5, wherein said upper and said lower limit of said predetermined proximity interval are equal. 9. A method of creating an associative database in one language, containing the stages at which - 52 007776 create a fund of documents, while said fund contains at least one document; choose the first word or phrase; find many occurrences of the first word or phrase; specify intervals in said fund, wherein said intervals occur depending on each of the plurality of occurrences of the first word or phrase; searching for said intervals in relation to common words or phrases, wherein said common words or phrases appear in a plurality of intervals; and associating the first word or phrase with common words or phrases based on the frequency of occurrence of common words or phrases within the above-mentioned intervals. INTERCONNECTIVE DATABASE one 2 3 4 Segments Segments Direct Frequencies after system A systems B ranked by frequency after subtraction frequencies subtraction 0a1 OY + OZ + O4 25 25 0a1 O9 + OYO nineteen nineteen Oa1 ϋΜ 35 10 Oa1 OY + OZ thirty 5 Oa2 0b2 + 0b6 twenty twenty Oa2 ϋΜ2 fifteen fifteen 0a2 0b2 25 5 OaZ + OaD b5 + b7 fifteen fifteen OaZ + Oa4 B5 25 10