JPH03244071A - Document proofreading back-up system - Google Patents

Abstract

PURPOSE:To reduce the scale of a document proofreading back-up system by deciding automatically a fact that the original or duplicate description is used again by mistake after making clear the correspondence between the original and duplicate descriptions during the production of a document and then using the duplicate or original descrip tion. CONSTITUTION:A document proofreading processing part 1 includes a morpheme analyzing part 4 which analyzes the input sentences in terms of a surface layer, a description correspondence detecting part 5 which detects the correspondence between the original and duplicate descriptions out of the input sentences, and a description use recognizing part 6 which recognizes the using state of each description based on the detected description correspondence data. Furthermore, a warning information generating part 7 produces the warning information to the improper use of description if detected based on the recognized using state of description. Thus the improper descriptions can be detected without using an original-duplicate description dictionary of a large scale if the possibility is confirmed for the irregular mixture of both original and duplicate descriptions when the sentences including many technical terms are produced. Thus, it is possible to obtain a document proofreading back-up system of a small scale that has high practicability.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to a document proofreading support system that points out inappropriate expressions when creating correct natural language sentences.

(Prior Art) In recent years, natural language processing technology using computers has improved, and various document processing systems have been developed.

In particular, document proofing support systems have been put into practical use that issue warnings and semi-automatically correct inappropriate expressions in documents created by users.

However, most of these require large-scale natural language processing dictionaries that can store tens of thousands to hundreds of thousands of words, making it difficult to miniaturize the system.

(Problem to be solved by the invention) Papers and explanatory texts in specialized fields are characterized by the fact that they contain many technical terms. Many of these terms are compound words made up of multiple words. When a compound word has a large number of constituent words, or when the length of each word is long, abbreviations and contractions are often used instead of the regular notation.

However, when creating sentences manually, as the size of the sentences increases, the use of regular notation and the use of sub-notation such as abbreviations and contractions may become irregularly mixed.

Irregular mixtures of regular notation and subnotation may be correct in terms of the meaning of the sentence, but inappropriate in terms of the form of expression.

By the way, the correspondence between the regular notation and the sub-notation can often be detected relatively easily from only the superficial character information when both are expressed in parallel.

Furthermore, by limiting the object of document processing to the superficial level as described above, the dictionary data required by the document proofreading support system can be considerably reduced.

This invention was made in view of the above-mentioned background, and its purpose is to clarify the correspondence between regular notation and sub-notation during writing, and then use the sub-notation (or regular notation). An object of the present invention is to provide a grammar proofing support system that automatically determines whether the correct notation (or subnotation) has been used again by mistake.

[Structure of the Invention] (Means for Solving the Problem) The grammar proofing support system of the present invention includes a means for inputting a sentence created in a natural language, and a means for determining the correspondence between the normal notation and the sub-notation in the input sentence. means for detecting an intended pair of descriptions and their description locations based on a predetermined algorithm that superficially uses character information in the corresponding sentences; and means for detecting each of the detected pair of descriptions in the sentences after the detection location and a means for determining misuse of the notation based on the fact that the search detects a mixture of a description corresponding to a subnotation and a description corresponding to a regular notation. do.

(Function) According to the present invention, it is possible to automatically determine that regular notation and sub-notation are mixed in a human-powered sentence. It can also warn you, which helps you correct mistakes that are often overlooked manually, increasing the efficiency of creating correct sentences.

In addition, for detecting a pair of descriptions that intend to correspond to a regular notation and a subnotation, and the place where they are written, from a human text,
Since it is based on a predetermined algorithm that superficially uses the character information in the text, there is no need for a dictionary in which the relationship between regular notation and sub-notation is registered in advance, and therefore it can be easily applied to small-scale text creation devices. Can be adopted.

(Example) Hereinafter, a document proofreading support system according to an example of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram showing a schematic configuration of an embodiment system.

The document proofreading processing unit 1, which forms the main part of this system, superficially analyzes the input sentence given via the human control device 2, processes the correspondence between the normal notation and the sub-notation, and based on the result, the user This system outputs warning information that is assumed to be the same through the output control device 3, and is numerically realized as a computer system equipped with document proofreading software I/ware.

The input control device 2 is for inputting document information to be processed as a string of characters from, for example, a magnetic storage medium or a keyboard.

The output control device 3 is for presenting the warning information outputted from the document proofing processing section 1 together with the input text using, for example, a CRT display device, or for printing out using a printer.

Now, the document proofreading processing unit 1 basically consists of a morphological analysis unit 4 that superficially analyzes an input sentence, an orthographic correspondence detection unit 5 that detects the correspondence between normal notation and subnotation from the input sentence, and an orthographic correspondence detection unit 5 that detects the correspondence between normal notation and subnotation from the input sentence. A notation usage recognition unit 6 recognizes the usage status of each notation based on the notation correspondence data detected by the notation usage recognition unit 5, and a notation usage recognition unit 6 that recognizes inappropriate usage based on the usage status recognized by the notation usage recognition unit 6. A warning information generating section 7 is provided to generate warning information in response to the warning information.

The morphological analysis unit 4 performs morphological analysis processing such as word segmentation of an input sentence based on simple dictionary data.

Note that the specific method of morphological analysis is not dependent on the present invention, and may utilize conventional techniques. Therefore, detailed explanation will be omitted.

The orthographic correspondence detection unit 5 detects a pair of words with a clear correspondence relationship from the input sentence based on the parallel expressions of the regular notation and the sub-notation, using an algorithm described below, and uses the result as the "regular notation" in the orthographic correspondence table 8. , "Subnotation yo" Register in the "first appearance style".

As shown in FIG. 2, the notation correspondence table 8 includes information such as "regular notation", "subnotation", "first appearance format", and "continuous use format".

When a notation registered in the notation correspondence table 8 appears in a subsequent sentence, the notation usage recognition unit 6 sequentially sets the used notation in the "continued usage style" of the notation correspondence table 8.

The warning information generation unit 7 monitors the contents of the notation correspondence table 8 set by the notation usage recognition unit 6, and when an inappropriate situation is discovered between the “first appearance format” and “continued usage format”, it Warning information is output to the output control device 3 along with the contents of the notation correspondence table.

FIG. 2 is a diagram showing the data structure of a notation correspondence table used by the document proofreading processing section of this embodiment.

Each element has the following meaning and is set and referenced as the document is processed.

The regular notation and the sub-notation are set when the orthographic correspondence detection unit detects a pair of words with a clear correspondence relationship from the input sentence based on the parallel expressions of the regular notation and the sub-notation.

Regular notation generally refers to a word expressed without abbreviating or abbreviating it, and sub-notation refers to a word expressed in a form different from the regular notation by abbreviation or abbreviation.

The parallel expression that is the basis for detecting the correspondence between regular notation and subnotation is, for example, “Lexical Functional
Grammar (LPG)", where a subnotation is written immediately after the regular notation by surrounding it with symbols such as parentheses, or conversely,
”+, pG (Lexfcal Function
There are cases where the regular notation is written immediately after the sub-notation, such as "Grammar)", by surrounding it with symbols such as parentheses.

In the first appearance format, when the correspondence between the regular notation and the sub-notation is detected, the notation of the surface layer 2 and the detection position are set at the same time. For example, the content of the notation is “Lexica
Functional Grammar (LPG)
”, “Main (sub)”, “LPG (Lexical)”
FunctionalGrammar)
For example, the content of the detection position may be
It is written as "page 10, line 23, column 7".

In the continuous usage format, the notation on the surface and the detection position are simultaneously set every time the regular notation or sub-notation of a word already registered in the notation correspondence table appears in a sentence. The content of the notation is set as "correct" if it is a regular notation, and "sub" if it is a subnotation. The description format of the detected position is the same as the first appearance format.

For example, the algorithm for detecting the correspondence between regular notation and sub-notation is similar to that for English.

The text to be processed is input as a word string, and the input word string is stored in a queue of sufficient size.

If ")" is found in the human word, the word string stored in the queue is searched for "(" backwards. If "(" is found, the word between the above ")" is searched backwards. Store the column as A.

Word string A is one word consisting of all capital letters and the number of characters is n.
In this case, the word string in the queue is searched for n more words, and that word string is designated as B. Word string A and word string B are candidates for parallel expression.

The first letters of each constituent word in word string B are consecutive, or if the string is in lower case, the upper case is,
If it matches the word string A, it is determined that the word strings A and B have a correspondence relationship of regular notation and subnotation, and that word string B is the regular notation and word string A is the subnotation. In some cases, the match may not be the first letter of all the constituent words of word string B, but a sequence of capital letters.

Conversely, if word string A is made up of multiple words,
If the word C immediately before "(" is a word with all capital letters, then the first letter of each constituent word in the word string is contiguous, or if the string is a lower case letter, it is capitalized, or the word string A is capitalized.) If a sequence of uppercase letters matches word string C, word strings A and C have a correspondence relationship of regular notation and subnotation, and word string A is in regular notation,
It is determined that the word string C is a sub-notation.

FIG. 3 is a flowchart showing a document proofreading procedure in the document proofreading processing section configured as described above.

This processing procedure is performed by inputting one word from the input control device 2 (step 301).
In step 2, the notation correspondence detection unit checks whether the correspondence between the regular notation and the subnotation is detected according to the above-mentioned algorithm.

When a new correspondence relationship between the regular notation and the sub-notation is detected, the process proceeds to step 303, where it is registered in the notation correspondence table 8 using the method described above. If no regular notation/subnotation correspondence is detected in step 302, the process proceeds to step 304.

In step 304, the spelling usage recognition unit 6 checks whether the input word matches a word (regular spelling or sub spelling) registered in the spelling correspondence table 8 so far. When a match with a word in the spelling correspondence table 8 is detected, data is added to the continuous usage pattern of the corresponding word (step 305).

Subsequently, in step 306, it is checked whether the current continuous use form data is the second or subsequent use form data. If it is the second or subsequent time, the process proceeds to step 307, where the warning information generation unit 7 compares the usage pattern with the previous continuous usage pattern and checks whether or not the warning information generation criteria are met.

The criteria for generating warning information here means, for example, that if the current continuous use form is "pressure" even though all the previous continuous use forms are "subj", it is inappropriate use. On the other hand, if the current continuous use form is ``secondary'' even though all previous continuous use forms are ``correct'', it will be judged as inappropriate use.

Thus, once inappropriate use is detected, step 3
At step 08, the warning information generating section 7 outputs the warning information together with the corresponding notation correspondence table data.

The processes of steps 301 to 308 are repeated until an input word is entered (step 309). The above are the basic processing procedures and changes in data in this embodiment.

Next, with reference to FIGS. 4 and 5, non-implementation 1 2 A specific example of processing in the example system will be explained.

Here, it is assumed that the input natural language sentence is an English sentence.

FIG. 4 is an example of an English sentence to be processed. This includes 'Lex' which is the regular notation and sub-notation of the same term.
ical Functjonal GraIntnar
"," and "LPG" are mixed together.

FIG. 5 is a diagram showing changes in data in the notation correspondence table when processing the document shown in FIG. 4.

In the document proofreading process, the notation correspondence detection unit 5 detects “Lexical FuncNonal G” in the 4th and 5th lines.
'Lexic' from the parallel expression of 'rammar(LPG)'
al Functional Grammar'“ and'
When detecting that “LPG” corresponds to the normal notation and sub-notation,
A new entry is made to the notation correspondence table, as shown in Figure 5(a).
The ``regular notation'', ``secondary notation'', and ``first appearance style'' are set as follows.

Next, when the notation usage recognition unit 6 detects 'LFG' in the 10th line, it matches the sub-notation of the word already registered in the notation correspondence table, so it is added to the notation correspondence table as shown in FIG. 5(b). “Sub” data is added to “Continued use form” [1].

Subsequently, when the notation usage recognition unit 6 detects "LPG" on the 13th line, it detects "LPG" on the 10th line as shown in FIG.
) in the notation correspondence table [2
] "Sub" data is added.

Up to this point, the warning information generation unit 7 has not generated any warning information, either because it is monitoring the notation correspondence table or because the warning information generation criteria are not met.

Next, the notation usage recognition unit 6 reads “LexicaI” on the 19th line.
When "Functional Grammar" is detected, "correct" data is added to the "continuous use format" [3 columns] in the notation correspondence table as shown in FIG. 5(d).

Here, the warning information generating section 7 inspects the notation correspondence table. "Continuous use format" Although [1] and [2] are "secondary", the same [3] are "normal", so the above-mentioned warning information generation criteria are met and the warning information is sent to the output control device 3. Output.

In this way, the system of this embodiment automatically detects the inappropriate expression in line 19, which would be more natural to write as "LPG" from the context, and makes it easy to create a valid English sentence. becomes.

It should be noted that the present invention is not limited to the second embodiment. The language to be processed is limited to English, but it can be implemented using the same method for Japanese and other languages.

When targeting Japanese, English terminology is often mixed, so we use parallel expressions in Japanese and English, such as ``Lexical Functional Grammar''.
When an English term is written in parentheses after a Japanese term, such as in [mmar) J or [Policy Functional Grammar (LPG) J], the Japanese term is written in the regular notation, and the English term is written in the secondary notation. It can also be treated as

In addition, as a processing method, you can manually edit the entire document, proofread the document online, and print out the warning results, or manually process the entire document in real time and input inappropriate expressions. An alarm can also sound at that point.

Furthermore, it is possible to have a function that not only warns about inappropriate expressions but also automatically corrects them to appropriate expressions.

Furthermore, the algorithm for detecting the correspondence between the regular notation and the sub-notation is not limited to the "-" symbol, but may recognize anything other than "0" as a mark of parallel expression.

Furthermore, in the above embodiment, since the first appearance style is only set and not referred to, it may be deleted from the notation correspondence table.

Conversely, it is also possible to use the first appearance format set at the beginning, and to standardize subsequent notations to the notation written outside the parentheses in the first appearance format.

In this case, in the example of FIG.
This will be modified to l Grammar.

Furthermore, after detecting the correspondence between the regular notation and the sub-notation for all documents, the recognition investigation of their usage status may be performed again from the beginning of the sentence.

In this way, it is possible to deal with the case where the regular notation or subnotation is used before the correspondence relationship is detected.

In addition, the present invention can be implemented with various modifications without departing from the spirit thereof.

[Effects of the Invention] 5 ] 6 As described above, according to the present invention, when creating a sentence containing many terminology, when there is a possibility that the regular notation and sub-notation are mixed irregularly, Inappropriate expressions can be detected without using a large-scale regular or sub-notation compatible dictionary, and a small but highly practical document proofreading support system can be realized.

[Brief explanation of drawings]

Fig. 1 is a block diagram showing a schematic configuration of the embodiment system, Fig. 2 is a diagram showing an example of data of a notation correspondence table used by the embodiment system, and Fig. 3 is a diagram showing the flow of document proofreading procedures in the embodiment system. 4 is an example of an English sentence to be processed by the embodiment system, and FIG. 5 is a diagram showing data changes in the notation correspondence table when processing the sentence in FIG. 4. ]...Document proofreading processing unit, 2...Input control device, 3...
- Output control device, 4... Morphological analysis unit, 5... Notation correspondence detection unit, 6... Notation usage recognition unit, 7... Warning information generation unit, 8... Notation correspondence table.

Claims (1)

[Claims] (1) A means for inputting a sentence created in a natural language, a pair of descriptions intended to replace the regular notation and the subnotation from the input sentence, and the written part by superficially converting the character information in the corresponding sentence. means for detecting based on a predetermined algorithm used; means for sequentially searching for each of the pair of detected descriptions in the text after the detection location; 1. A grammar proofing support system comprising means for determining misuse of a notation based on detection of a mixture with a corresponding description.