JPH0481971A - Machine translation apparatus - Google Patents

Abstract

PURPOSE:To efficiently execute the translation by retrieving a generation dictionary of a translation result sentence, when the language of the translation result sentence is obtained in an input translation object sentence, retrieving the generation dictionary by deleting one character each from the ending of the language, when the language corresponding to a keyword of the generation dictionary does not exist, and also, retrieving a derived form of the keyword, as well. CONSTITUTION:A machine translation apparatus 10 is provided with a data processing part 18 being a computer main body to which an input part 12, a display part 14, and a printing part 16 are connected. In such a state, when a second language is contained in an inputted translation object sentence, a generation dictionary 28 is retrieved, and dictionary information is given to this second language. In this case, when the second language does not exist in a keyword of the generation dictionary 28, the generation dictionary 28 is retrieved successively by deleting one character each from the ending of second language, and also, by retrieving a derived form of the keyboard, as well, appropriate dictionary information is given to second language. In such a way, the efficiency of a translation is improved.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a machine translation machine that translates a translation target sentence in a first language, including a second language, into a translation result sentence in a second language.

More specifically, when the language constituting the sentence to be translated is identified as the first language, an analysis dictionary for the first language is searched, while when the language is identified as the second language, the second language is searched. By searching the generation dictionary, necessary dictionary information can be obtained and a target sentence to be translated that includes the second language can be generated as a desired translation result sentence.

[Prior Art] Recently, there has been a desire to improve the efficiency of translation into different languages, and machine translation machines based on computer systems are being used.

In a translation target sentence applied to such a machine translator, the language of the translation result sentence may be included in the translation target sentence. For example, when translating a Japanese sentence that is a translation target sentence into an English sentence that is a translation result sentence, there are cases where it is desired to insert English words into the Japanese sentence and execute the translation. In this case, the English words contained in the Japanese sentence will be treated as unknown words, or
Alternatively, the translation is performed as a simple alphabetic string.

[Problem to be solved by the invention]

In this way, in the above-mentioned conventional machine translation machine,
English words contained in the Japanese sentence to be translated are processed as unknown words or simply as alphabetic character strings to generate translated sentences, and these character strings are not stored in dictionaries. is not searched and necessary dictionary information cannot be added. Therefore, the user is required to perform post-editing, which is more labor-intensive, on the generated translation result sentence, resulting in a disadvantage that translation efficiency is reduced.

The present invention has been made in view of the above points, and an object of the present invention is to perform post-editing etc. on the translation target sentence when generating a first language translation target sentence including a second language into a second language translation target sentence. Our goal is to provide an excellent machine translator that can translate efficiently without any hassle.

[Means to solve the problem]

In order to solve the above problems, the machine translator of the present invention has the following features:
As shown in the claim correspondence diagram in Figure 1, there is provided a machine translation machine that generates a translation target sentence in a first language including a second language into a translation result sentence in a second language, the identification means (A) comprising: , distinguish between the first language and the second language.

Further, in the first search means (B), when the language is identified as the first language, the analytic dictionary of the first language is searched.

Further, in the second search means (C), when the second language is identified, the second language generating dictionary is searched.

Further, in the re-search means (D), the above-mentioned identification means (A)
When the language identified as the second language does not match the headword of the generated dictionary, the generated dictionary is sequentially searched for a language obtained by deleting each letter from the end of the language.

Furthermore, the re-searching means (D) is characterized in that it refers to the derived form of the headword in the generated dictionary.

[Effect]

In the machine translator of the present invention configured as described above, when a second language is included in the input sentence to be translated, the generated dictionary is searched and dictionary information is added to the second language.

In this case, if the headword in the generated dictionary does not exist in the second language, the generated dictionary is sequentially searched by deleting one character at a time from the end of the word in the second language, and the derived form of the headword is also searched. This makes it possible to provide appropriate dictionary information to the second language.

〔Example〕

Next, embodiments of a machine translator according to the present invention will be described in detail below with reference to the accompanying drawings.

FIG. 2 is a configuration diagram, FIG. 3 is a flowchart corresponding to a control operation program, and FIG. 4 is a diagram showing the configuration contents of a generation dictionary.

First, the configuration will be explained. In FIG. 2, 10 is a machine translator such as a computer system, and this machine translator IO includes an input section 12 consisting of a keyboard etc., a display section 14 such as a CRT, and a printing section 16 such as a scanning printer. is provided. Furthermore, the machine translator IO further includes an input unit 12,
It has a data processing section 18 which is a computer main body to which a display section 14 and a printing section 16 are connected.

The data processing section 18 includes an editing section 20 and a translation section 22, which are function combining means, as well as an analysis dictionary 24, a conversion dictionary 26, and a generation dictionary 28 stored on a hard disk or optical disk.

The translation unit 22 includes a morphological analysis unit 30, which is a functional combination means related to a well-known transfer method, and a syntactic analysis unit 32.
Syntactic converter 34, syntactic generator 36, and morpheme generator 3
8 is provided.

Here, the basic operation in the above configuration will be explained.

The translation target sentence in the first language input from the input unit 12 is edited into an appropriate form by the editing unit 20 and then input to the translation unit 22 . The sentence to be translated inputted to the translation unit 22 is divided into words by the morphological analysis unit 30 with reference to the analysis dictionary 24 whose heading is ``first eaves'', and the parts of speech and conjugations are clarified. Next, the parsing unit 32 also uses the parsing dictionary 24.
By referring to , necessary dictionary information is added and the syntax is clarified.

After that, the translation target sentence is sent to the syntactic conversion unit 34, and bilingual translation is performed with reference to the conversion dictionary 26 that indicates correspondence from the first language to the second language, and the translation target sentence is converted from the translation target sentence in the second language. is converted to

Then, the translation target sentence converted into a translation target sentence by the syntax conversion unit 34 is converted into a translation target sentence by the syntax generation unit 36 based on the dictionary information and generation processing of the generation dictionary 28 with the second language as the heading.
Generated with correct syntax. Thereafter, the morpheme generation unit 38 completes the process and generates a final translation result sentence. Further, the translation result sentence is freely edited in the editing section 20 and displayed as an image on the display section 14, or output as a hard copy from the printing section 16.

Next, the operation of the machine translator 10 described above to generate a translated sentence when the Japanese sentence, which is the sentence to be translated, includes English words will be explained with reference to the flowchart of FIG.

First, in step 100 (hereinafter referred to only as S), for example, a sentence to be translated [I ate a lot of apples. ” is input from the input unit 12.

Next, in the 5IOI, the morphological analysis unit 30 of the translation unit 22 performs morphological analysis, and divides the sentence to be translated into words such as “I/ate/ate a lot of /apples/.J”.

Next, in 5102, it is determined whether the sentence to be translated includes an alphabetic character string. If no,
Proceeding to step 5103, dictionary information is added by referring only to the Japanese analysis dictionary 24. If Yes, it is assumed that the sentence to be translated contains an alphabetic character string, and 5
Proceeding to step 104, the Japanese analysis dictionary 24 and the English generation dictionary 28 are searched.

Next, at 5105, the English word "applesJ"
The generated dictionary 28 is searched to determine whether there is a headword.

In this case, since the English word [applesJ is plural, there is no corresponding entry in the generated dictionary 28. If there is a corresponding headword here, proceed to 5IIO. Furthermore, in 5106, proceed in case of No, the English word r
Delete the English character rsJ from the end of applesJ,
Thereafter, in 3107, the entry word in the generation dictionary 28 is searched again. In this case, by deleting one character at the end of the word, it is possible to search for the same form as the headword in the generation dictionary 28.

On the other hand, some English words include the above word "applesJ"
In addition to words that change regularly, such as the English word ``rdictionary'', there are also words that change their endings irregularly, such as the English word ``rdictionary''.For such words, even if you delete one letter from the end of the word, you will not be able to get the desired headword. Not only can you search for other meanings of the English word ``rdiction''
Sometimes it gets searched.

On the other hand, in this embodiment, as shown in FIG.
In the case of verbs, it also remembers the conjugation form, etc.

Therefore, the English word rdictionariesJ
In the case of 5105 above, it is determined that it is not the same as the headword,
That is, in the case of NO, one word at the end of the word is further deleted in 5107. In this case as well, after it is determined that the NO is not the same as the headword, the English word [di
The English word d1cti is a derivative of ctionaryJ.
The English word rdi as identical to onarieJ
ctiOnar)FJ is searched. If the answer is No, the process proceeds to 5109, and if the answer is Yes, the process proceeds to 5IIO.

In this way, even in Sl 05.3107.5108, if the corresponding headword is not searched, 51
09, it will be processed as an unknown word. That is, the translation result sentence generated after this is processed into a normal desired translated sentence by post-editing.

The headword is searched as described above, and 5IIO
Dictionary information such as part of speech is added.

Next, the syntax is analyzed in step 511, and a translation target sentence is generated with reference to the conversion dictionary 26 in step 5112.

Subsequently, in 5l13, a correct syntax is generated from the translation target sentence, and in 5114, a desired translation result sentence is generated and displayed on the display section 14, and the program related to the processing procedure of the process ends.

As described above, when the translation target sentence includes the language of the translation result sentence, the generation dictionary 28 is also searched, and dictionary information is thereby added. In this case, even if the language does not match the headword in the generation dictionary 28, a search is performed based on the derived form, so it is possible to generate an extremely accurate translation result sentence.

In the above example, the first language is Japanese and the second language is English. The present invention also includes application to obtain similar effects.

C. Effects of the invention] As can be understood from the above explanation, according to the machine translator of the present invention, when the input translation target sentence includes the language of the translation result sentence, the translation result sentence is The generated dictionary is searched, and if there is no language corresponding to the headword in the generated dictionary, the generated dictionary is searched by deleting one character from the end of the word in the language mentioned above, and a search is performed with reference to the derived form of the headword. This eliminates the need for time-consuming post-editing, and has the advantage of efficiently generating translated sentences.

[Brief explanation of the drawing]

Fig. 1 is a claim correspondence diagram of the machine translation machine of the present invention, Fig. 2 is a configuration diagram showing an embodiment of the machine translation machine of the present invention, and Fig. 3 is a control operation of the embodiment shown in Fig. 2. The flowchart corresponding to the program shown in FIG. 4 is a diagram showing the configuration contents of the generated dictionary. 12... Input section, 14... Display section, 16...
Printing department, 18... Data processing department, 20... Editorial department, 22... Translation department, 24... Analysis dictionary, 26.
...Conversion dictionary, 28...Generation dictionary, 30...Morphological analysis section, 32...Syntax analysis section, 34...Syntax conversion section, 36...Syntax generation section, 38...Morphological generation Department.

Claims (3)

[Claims] (1) A machine translation machine that generates a translation target sentence in a first language including a second language into a translation result sentence in a second language, an identification means for identifying the first language and the second language; a first search means for searching an analytical dictionary for the first language when the language is identified as the first language by the identification means; and a generating dictionary for the second language when the language is identified as the second language by the identification means; A machine translation device characterized by comprising a second search means for searching. (2) The second search means, when the language identified as the second language by the identification means does not match the headword of the generation dictionary,
2. The machine translation machine according to claim 1, further comprising re-search means for sequentially searching the generated dictionary for a language formed by deleting one character from the end of a language. (3) The machine translator according to claim 2, wherein the generating dictionary includes a derived form of the headword, and the re-search means refers to the derived form.