JP2002032369A - Dictionary-preparing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To realize a dictionary-preparing device capable of preparing a parallel translation type dictionary even from text in a single language. SOLUTION: A conversion table 33 describes a target language part of speech, which corresponds to the part of speech of the word string of the original language and a conversion rule from the original language word string into a target language word string. An extraction processing part 21 extracts a word string from the text in the original language. A part of the speech attaching part 22 attaches a part of the speech to the word string. A conversion processing part 23 converts the word string in the original language into a word string in the target language, on the basis of the attached part of speech according to the conversion rule by referring to the table 33 and outputs the word string in the target language as dictionary data.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a dictionary creating apparatus for automatically creating a dictionary used in a machine translation system or the like.

[0002]

2. Description of the Related Art With the spread of the Internet and an increase in the number of personal computer users, a machine translation system for translating a text described in one language into another language has rapidly spread. However, in a general machine translation system, sentences in various fields cannot be appropriately translated only by a dictionary provided by the system, and thus it is essential to construct a dictionary for each user or each field. Creating such a dictionary requires a great deal of effort. Therefore, there has been proposed a method of automatically creating a machine translation dictionary from a translation target itself or text data in the same field as the translation target.

As such a prior art, Japanese Patent Laid-Open No.
There is a method disclosed in Japanese Patent Publication No. 269222, “Dictionary creation support device in machine translation”. In this method, a character string in Katakana notation is extracted from Japanese text data, a set of English spelling candidates is obtained from the character string in Katakana notation using a conversion table, and the set of English spelling candidates and the English text data are extracted. By collating, the translation is automatically given.

[0004]

However, the above prior art has the following problems. First, text data in two languages (Japanese and English in this case) is always required. Second, since English translation candidate words are estimated from Japanese words in katakana notation, Japanese words without katakana notation cannot be registered. Third, if there is no word that matches the set of spelling candidates in the English text data, a translated word cannot be obtained.

As described above, in the prior art, there is a problem that a bilingual dictionary cannot be created from a monolingual text.

[0006]

The present invention employs the following structure to solve the above-mentioned problems. <Structure 1> An extraction processing unit that extracts a word string from a text in the source language, a part-of-speech providing unit that adds the part of speech of the word string extracted by the extraction processing unit, and a target language corresponding to the part of speech of the word string in the source language A part-of-speech unit refers to the conversion table that describes the part of speech and a conversion rule from the source language word string to the target language word string, and the conversion rule of the conversion table for the word string extracted by the extraction processing unit. A conversion device for converting a word string of a target language into a word string of a target language based on a part of speech of an assigned word string, and outputting the word string of the target language as dictionary data of a source language.

<Structure 2> In the dictionary creating apparatus described in Structure 1, the conversion rule of the conversion table is constituted by contents obtained by replacing words in the source language and functional words in the target language. Dictionary creation device.

<Structure 3> The dictionary creating apparatus according to Structure 2, wherein the contents of the source language words replaced are rules using the source language words as they are.

<Structure 4> In the dictionary creating apparatus according to Structure 2, the contents of the source language words replaced are rules for expressing the source language words in phonograms.

[0010]

DESCRIPTION OF THE PREFERRED EMBODIMENTS << Summary of the Invention >> When a technical document or the like is translated into Japanese, which is a target language, the terminology is translated into the original language in order to avoid mistranslation and eliminate ambiguity. In many cases, the words in the original language are used as they are, or the translated words are obtained by simply replacing the words in the source language with katakana. Therefore, the present invention extracts a word or word string from text data in one language, and converts the word or word string in the source language into the target language by using a conversion table for the target language, thereby automatically converting the translated word. And a dictionary creation device capable of creating a dictionary.

Hereinafter, embodiments of the present invention will be described in detail using specific examples. << Specific Example >><Configuration> FIG. 1 is a configuration diagram showing a specific example of a dictionary creation device of the present invention. The illustrated device includes an input / output device 1 and a processing device 2
And a storage device 3. The input / output device 1 includes input means such as a keyboard, a mouse, and a file for inputting text data and various operation commands, and output means such as a CRT for displaying a processing process and an output file. The storage device 3 is composed of a magnetic disk device, a semiconductor memory, or the like, and has a text file 31 for storing texts of source language sentences, a work file 32 for storing processing results of each stage, and a source language to a target language. It has a conversion table 33 in which rules for conversion into Japanese are described, and a dictionary file 34 for storing the created dictionary.

The work file 32 is shown in FIGS.
, A source language part-of-speech storage unit 321 storing source language parts of speech, a source language word storage unit 322 storing source language word strings, and a Japanese part-of-speech storage unit 32 storing Japanese parts of speech.
3 and a Japanese word storage unit 32 for storing a Japanese word string
4. As shown in FIG. 6, which will be described later, the conversion table 33 includes a source language part of speech storage unit 331 that stores source language parts of speech, a Japanese part of speech storage unit 332 that stores Japanese parts of speech, and a conversion rule for Japanese. Is stored in the conversion rule storage unit 333. Dictionary file 34
7, a source language part-of-speech storage unit 341 storing source language parts of speech, a source language word storage unit 342 storing source language word strings, and a Japanese language part storing Japanese parts of speech, as shown in FIG. It comprises a part-of-speech storage unit 343 and a Japanese word storage unit 344 that stores a Japanese word string.

The processing device 2 includes an arithmetic unit, a memory, a control unit, and the like, and has a function of executing a translation pattern creation process. The processing device 2 includes an extraction processing unit 21, a part of speech giving unit 22, and a conversion processing unit 23. The extraction processing unit 21 performs a process of extracting words and word strings (hereinafter, referred to as registration candidates) to be registered in the dictionary from the source language text file 31 and storing the words and word strings in the source language word storage unit 322 of the work file 32. It has. The part-of-speech providing unit 22 stores the source language word storage unit 322
Has a function of reading the contents stored in the Japanese part-of-speech line by line, estimating the part of speech, and storing it in the Japanese part of speech storage unit 323.
The conversion processing unit 23 reads a set of a word string and a part of speech of a registration candidate stored in the work file 32 one line at a time, converts the set into a translated word based on the conversion rule of the corresponding part of speech in the conversion table 33, It has the function of storing.

<Operation> FIG. 2 is a flowchart showing the flow of processing in a specific example. FIG. 3 is a flowchart showing details of step S5 in the flowchart of FIG. Hereinafter, in this specific example, the process of each process when an English-Japanese bilingual dictionary is created from English text will be described using a specific example.

[Step S1] An English text is input from the input / output device 1, and the extraction processing unit 21 stores it in the text file 31. The text may be input directly from the keyboard or a method of specifying a file. [Step S2] The extraction processing unit 21 reads the text file 31, and extracts a word string that is a candidate for dictionary registration. As a method of extracting a word string, there is a method of extracting a word string (for example, a noun phrase or the like) having a part of speech or a syntactic structure to be extracted by performing a morphological analysis or a syntactic analysis, or a method of obtaining an appearance frequency of a word string in a text, There is a method of extracting a high word string, and the like.

[Step S3] The extraction processing section 21 stores the extracted word string in the source language word storage section 3 of the work file 32.
22. FIG. 4 is an explanatory diagram showing the contents of the work file 32 after the extraction processing has been performed. In this example, “URL prefix”, “map”, “a
"dditional" is stored in the source language word storage unit 322.

[Step S4] The part of speech assigning section 22 reads the contents of the source language word storage section 322 of the work file 32 line by line, and estimates the part of speech of the word string. General morphological analysis and syntax analysis are used for the part-of-speech assignment. The estimated part of speech is stored in the source language part of speech storage unit 321 of the work file 32. FIG. 5 shows the contents of the work file 32 after the part-of-speech giving process has been performed. As shown, “URL
noun for "prefix", verb for "map", "additional"
Is given an adjective.

[Step S5] The conversion processing section 23 reads the contents of the source language word storage section 322 and the source language part-of-speech storage section 321 of the work file 32 line by line, and converts the contents with the part of speech read from the source language part-of-speech storage section 321. The part of speech in the source language part of speech storage unit 331 of the table 33 is collated. As a result of matching,
If there is a match, the word string read from the source language word storage unit 322 of the work file 32 is converted into Japanese according to the corresponding conversion rule.

FIG. 6 is an explanatory diagram showing an example of the conversion table 33. Here, “***” in the conversion rule storage unit 333
Indicates that the English character string is directly substituted for the translated word, but it is also possible to substitute the result of converting from English to katakana, or to translate the English using a machine translation system The result may be substituted. As described above, the conversion rule storage unit 333 of the conversion table 33 is used.
Is composed of the contents in which words in the source language are replaced, and functional words in Japanese, that is, words indicating the grammatical function of the part of speech, such as "do" in the case of a verb.

Next, the processing of the conversion processing unit 23 in step S5 will be described in detail with reference to FIG. [Step S51] It is checked whether the data of the work file 32 is not completed. If it is over, the process is terminated. Otherwise, the process proceeds to step S52. [Step S52] 1 data from the work file 32
Read the line. [Step S53] The part of speech in the source language part of speech storage 321 of the work file 32 is compared with the part of speech in the source language part of speech storage 331 of the conversion table 33. [Steps S54 to S55] If there is a matching part of speech,
A translation word is generated according to the conversion rule corresponding to the matching source language part of speech in the conversion table 33. [Step S56] The Japanese part of speech corresponding to the matching source language part of speech in the conversion table 33 and the generated translated word are stored in the Japanese part of speech storage unit 323 and the Japanese word storage unit 324 of the work file 32, respectively. [Step S57] If there is no matching part of speech, error processing is performed, and the process returns to step S51. The above is the detailed operation of step S5 in FIG.

Returning to FIG. 2, the processing of a specific example will be described. [Step S6] In the dictionary file 34, the registration candidates stored in the work file 32, their translations, and their parts of speech are stored.

Next, the flow of the conversion process will be specifically described using the data of the work file 32 of FIG. 5 and the data of the conversion table 33 of FIG. [Step S52] First, one line of data is read from the work file 32. Part of speech "noun" and English word string "URL prefix"
Is read. [Step S53] Next, the read part of speech “noun” is compared with the part of speech in the source language part of speech storage unit 331 of the conversion table 33. [Steps S54 to S55] Since "noun" is present in the conversion table 33, a translated word is generated according to the corresponding conversion rule "***". Since “***” is to replace the English word string as it is, the English word string becomes the translation word as it is. [Step S56] The Japanese part of speech “noun” corresponding to the English part of speech “noun” and the translated word are stored in the Japanese part of speech storage unit 323 and the Japanese word storage unit 324 of the work file 32, respectively. Returning to step S51, the same processing is repeated for the data in the second and third rows, and when the processing is completed for all the data, the contents stored in the work file 32 are stored in the dictionary file 34. I do.

FIG. 7 is an explanatory diagram showing the contents of dictionary data generated from the data of FIGS. 5 and 6. As illustrated, the words “URL prefix” and “ma” in the source language word storage unit 342 are used.
source language part of speech storage unit 341 corresponding to “p” “additional”
Stores “noun”, “verb” and “adjective”, and the Japanese part of speech storage unit 343 stores “noun”, “verb” and “adjective verb”.
Is stored in the Japanese word storage unit 344, and “URL pref
ix "," map "and" additional "are stored.

FIG. 8 is an explanatory diagram showing changes in the translation result of the machine translation system when the dictionary data of FIG. 7 is not used and when it is used. Here, as the input sentence 81,
An example in which “Type the URL prefix you want to map.” Is input is shown. The result 82 of translating the input sentence 81 only with the standard dictionary is "Type the URL prefix you want to map." In addition, the standard dictionary and FIG.
In the result 83 translated using the dictionary of
Type the URL prefix you want. Here, for example, in the case of "map" in the input sentence 81, there are translation words such as "make a map", "position", and "find out" in a general English-Japanese dictionary. Is not appropriate. In this case, “prefix” is a group of words “URL prefix”. This is, for example, http://www.abcd.com/English/index.ht
ml other than the file name “index.html” of the URL, ie, www.abcd.com/English, English, or htt
Because it points to p: //www.abcd.com, "U
The "RL prefix" is not accurate as the meaning of "URL prefix". Therefore, in such a case, in order to avoid erroneous translation, it is desirable to use the source language words as they are.

In the above example, the conversion from English to Japanese has been described, but the present invention can be applied to translation in any two languages. For example, when converting Japanese to English, a translated word is generated using the following conversion table. FIG. 9 is an explanatory diagram illustrating an example of the conversion table when the source language is Japanese. As the translated word, a Japanese word may be generated as it is, or a process of replacing the reading with a Roman character may be added. FIG. 10 is an explanatory diagram of the generated dictionary data. In this example, “yokozuna” and “ozeki” are registered as English words for Japanese nouns “Yokozuna” and “Ozeki”.

FIG. 11 is an explanatory diagram showing changes in the translation result of the machine translation system when the dictionary data of FIG. 10 is not used and when it is used. Here, an example is shown in which “Yokozuna has won Ozeki” has been input as the input sentence 91. In the result 92 of translating the input sentence 91 only with the standard dictionary, “A grand champion sumo wrestler defeated a su
mo wrestler of the second highest rank. "
In addition, as a result of translation using the standard dictionary and the dictionary of FIG.
In 3, it becomes “Yokozuna” defeated “ozeki”. Thus, by spelling close to the original sound and adding quotation marks “”, such as “yokozuna” and “ozeki”,
You can see that it is a foreign language other than English. Further, in the translation process, it is possible to prevent erroneous translation, for example, in which “Yokozuna” is translated by dividing into “Yokozuna” and “Tsunagi”.

<Effects> As described above, according to this specific example, the source language word sequence is converted to the source language word sequence using the conversion table that describes the target language part of speech and the conversion rule corresponding to the part of speech of the source language word sequence. Is converted to a target language word string based on the conversion rules and output as the source language dictionary data, so that a bilingual dictionary used for machine translation and the like is automatically created from monolingual text. It becomes possible. Further, when the target language is, for example, Japanese, it can be registered even if it is not necessarily a word written in katakana. Furthermore, since the conversion rule based on the part of speech of the word is used, there is an effect that it is not necessary to prepare a spelling candidate for obtaining a translated word from the text in the source language in advance.

In the above example, the conversion rule may be written in katakana notation. However, the conversion rule is not limited to katakana notation. Good.

[Brief description of the drawings]

FIG. 1 is a configuration diagram showing a specific example of a dictionary creation device of the present invention.

FIG. 2 is a flowchart showing an overall process in a specific example of the dictionary creation device of the present invention.

FIG. 3 is a flowchart illustrating details of conversion processing in a specific example of the dictionary creation device of the present invention.

FIG. 4 is an explanatory diagram showing the contents of a work file after an extraction process has been performed in a specific example of the dictionary creation device of the present invention.

FIG. 5 is an explanatory diagram showing the contents of a work file after a part-of-speech giving process is performed in a specific example of the dictionary creation device of the present invention.

FIG. 6 is an explanatory diagram showing an example of a conversion table in a specific example of the dictionary creation device of the present invention.

FIG. 7 is an explanatory diagram showing the contents of generated dictionary data in a specific example of the dictionary creation device of the present invention.

FIG. 8 is an explanatory diagram showing a change in a translation result of the machine translation system in a case where dictionary data of a specific example is not used and a case where dictionary data is used.

FIG. 9 is an explanatory diagram showing an example of a conversion table when the source language is Japanese.

FIG. 10 is an explanatory diagram of generated dictionary data when the source language is Japanese.

FIG. 11 is an explanatory diagram showing a change in a translation result of the machine translation system in a case where dictionary data of a specific example is not used and a case where dictionary data is used when a source language is Japanese.

[Explanation of symbols]

 21 extraction processing unit 22 part of speech assigning unit 23 conversion processing unit 31 text file 33 conversion table 34 dictionary file

Claims (4)

[Claims] 1. An extraction processing unit for extracting a word string from a text in a source language, a part of speech providing unit for adding a part of speech of the word string extracted by the extraction processing unit, and a part of speech of a word string in the source language. A conversion table describing target language part of speech, a conversion rule from the source language word string to a target language word string, and a conversion rule of the conversion table for the word string extracted by the extraction processing unit. And a conversion processing unit that converts the word string of the target language into a word string of the target language based on the part of speech of the word string given by the part of speech giving unit, and outputs the word string of the target language as dictionary data of the source language. Dictionary creation device. 2. The dictionary creation apparatus according to claim 1, wherein the conversion rule of the conversion table is composed of contents obtained by replacing words in a source language and function words in a target language. Creating device. 3. The dictionary creation device according to claim 2, wherein the contents of the replacement of the source language words are rules using the source language words as they are. 4. The dictionary creation device according to claim 2, wherein the contents of the replacement of the source language words are rules for expressing the source language words in phonograms.