JP2012133689A - Information processing apparatus, method and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically create a new bilingual dictionary, on the basis of more various bilingual dictionaries, with appropriate precision.SOLUTION: For both first and second languages for which a new bilingual dictionary is to be created, the same or similar material texts are prepared, the same or similar words are extracted between translated sentences translated into a third language, and words in both the first and second languages corresponding to the extracted word are made correspondent as a bilingual dictionary. Thus, if there are first→third and second→third bilingual dictionaries with appropriate precision, even in the case where there is no third→second bilingual dictionary or precision is reduced, on the basis of the similarity of words translated into the third language, the first→second bilingual dictionary can be automatically created with appropriate precision.

Description

  The present invention relates to machine translation.

  2. Description of the Related Art Conventionally, a so-called machine translation service has been provided in which an input original text is converted into another language and displayed on an Internet website or the like. As a device to improve the accuracy of this type of machine translation, there is a proposal to preferentially adopt a frequently used expression as a search result by inputting candidate translations based on the original sentence into a web search engine. (For example, refer to Patent Document 1).

  By the way, conventionally, in machine translation, a bilingual corpus, which is data in which a word in a certain language is associated with a translation in another language (hereinafter also referred to as “a bilingual dictionary”, is also referred to as “a bilingual dictionary” hereinafter). is required. The bilingual dictionary is not limited to the narrow-sense bilingual dictionary in which the translated words are associated with each other as described above. In addition, the bilingual dictionary functions as a system including bilingual rules, examples, and probability models (data and data structure). , Device, system, program, etc.). And since such a bilingual dictionary was created manually, the burden of labor etc. was great, and the automation was potentially desired.

JP 2007-87157 A

Kishore Papineni, “BLUE: a Method for Automatic Evaluation of Machine Translation,” ACL pp.311-318 (2002)

  In this case, it may be possible to create a bilingual dictionary using a new language combination using an existing bilingual dictionary. For example, if there are two types of bilingual dictionaries of Japanese (English), English (English), and Chinese (Medium): Japanese → English, English → Chinese, combine both with English as a relay, and say Japanese → Chinese. A bilingual dictionary can be created.

  However, since there is a direction in the bilingual dictionary, a bilingual dictionary in the direction of Japanese → English and English → Central is required as a basis for newly creating a bilingual dictionary of Japanese → China. Therefore, if there is a bilingual dictionary for the first half of the correspondence, the bilingual dictionary for English → English, but there is no bilingual bilingual dictionary for the second half, or if the accuracy is not sufficient, the bilingual dictionary for the reverse direction, middle → english, is used. However, there is a problem that a new bilingual dictionary based on a combination of bilingual dictionaries cannot be automatically generated with appropriate accuracies.

  In view of the above problems, an object of the present invention is to automatically generate a new bilingual dictionary with appropriate accuracy based on a more diverse bilingual dictionary.

  Based on the above object, one aspect (1) of the present invention provides a new bilingual dictionary for translating a first language into a second language, and a bilingual dictionary according to the first language and the second language. A first bilingual dictionary for translating the first language into a third language, and a second for translating the second language into the third language. Storage means for the new bilingual dictionary, a first material text in the first language, and content identical to or related to the first material text, Material acquisition means for acquiring the second material text in the language, and the acquired first material text is translated into the first translation in the third language using the first bilingual dictionary Use the second bilingual dictionary for the acquired second material text Translation means for translating into a second translation in the third language, and similarity for extracting words having a predetermined similarity between the translated first translation and the second translation as similar words A word extraction means, a word in the first language corresponding to the extracted similar word in the first material text, and the second corresponding to the extracted similar word in the second material text It is characterized by comprising: bilingual word extracting means for extracting a language word as a set of parallel translation words; and dictionary registration means for registering the extracted parallel word set in the new bilingual dictionary.

  According to another aspect (8) of the present invention, the above aspect is captured by a method category, and a new bilingual dictionary for translating a first language into a second language using a computer is provided. In an information processing method generated based on a bilingual dictionary related to one language and a second language, a computer translates the first language into a third language, the first bilingual dictionary, and the second bilingual dictionary Storage means for translating the second language into the third language and the new bilingual dictionary, and the computer implements the first language in the first language. A material acquisition process for acquiring a material text and a second material text in the second language that is the same as or related to the first material text; The first bilingual dictionary of material text A translation process that translates the second material text into the second translation in the third language using the second bilingual dictionary using the second translation text A similar word extraction process in which a computer extracts a word having a predetermined similarity between the translated first translated sentence and the second translated sentence as a similar word, and the computer extracts the similar A set of parallel translations of a word in the first language corresponding to a word in the first material text and a word in the second language corresponding to the extracted similar word in the second material text And a bilingual word extraction process for extracting the pair of bilingual words extracted in the new bilingual dictionary.

  Another aspect (9) of the present invention is a new bilingual dictionary for translating the first language into the second language by controlling the computer by capturing the above aspect in the category of a computer program. Is an information processing program that is generated based on a bilingual dictionary related to the first language and the second language, the program for translating the first language into a third language on a computer Storage means for the first bilingual dictionary, the second bilingual dictionary for translating the second language into the third language, and the new bilingual dictionary are realized, and the computer The first material text in the first language and the second material text in the second language that is the same as or related to the first material text, and obtained by the computer in front The first material text is translated into the first translation in the third language using the first bilingual dictionary, and the acquired second material text is translated using the second bilingual dictionary A second translation in a third language is translated, and the computer is caused to extract words having a predetermined similarity between the translated first translation and the second translation as similar words, and the computer A word in the first language corresponding to the extracted similar word in the first material text; a word in the second language corresponding to the extracted similar word in the second material text; Are extracted as a set of parallel translation words, and the computer is caused to register the extracted pair of parallel translation words in the new parallel translation dictionary.

  In this way, the same or similar material text is prepared for both the first and second languages for which a new bilingual dictionary is to be created, and the same and similar words are extracted between the translated sentences translated into the third language. Then, words in both the first and second languages corresponding to the extracted words are associated with each other to form a bilingual dictionary. As a result, if there is a first → third and second → third bilingual dictionary with appropriate accuracy, even if there is no third → second bilingual dictionary or the accuracy is inferior, it is translated into the third language. Based on the similarity between words, a first-to-second bilingual dictionary can be automatically created with appropriate accuracy.

  According to another aspect (2) of the present invention, in any one of the above aspects, the material acquisition unit includes a plurality of web pages in the first language based on one keyword from a predetermined web search system. A second web search result including a plurality of web pages in the second language based on the keywords from a predetermined web search system, and a web page in the first language Among the second language web pages, a first language web page and a second language web page having predetermined common features are respectively designated as the first material text and the second material text. It is characterized by acquiring as.

  In this way, each page selected based on commonality from the web page group of the first language and the web page group of the second language searched and hit based on the same keyword in the web search system is translated. By obtaining the first material text and the second material text as the target, the probability that similar words and corresponding words will appear in similar usage and context in each document and translation will be high, The accuracy of the bilingual dictionary based on it is improved.

  In another aspect (3) of the present invention, in any of the above aspects, the predetermined common feature is that the web page in the first language and the web page in the second language are the same or similar to each other. It is characterized by including an image.

  In this way, by acquiring a set of the first material text and the second material text based on the commonality of the images, the image often specifies a topic in a straightforward manner. Therefore, it is possible to easily create an accurate bilingual dictionary based on.

  In another aspect (4) of the present invention, in any one of the above aspects, the predetermined common feature is that the relative rank of the click rate in the first web search result or the second web search result is close. It is characterized by being.

  In this way, a set of the first material text and the second material text having the same relative click rate when the web search result of the first language or the second language is displayed on the access source terminal is acquired. By doing so, it is possible to easily create an accurate bilingual dictionary based on a set of documents having similar positions such as popularity and attention.

  In another aspect (5) of the present invention, in any one of the above aspects, the material acquisition unit acquires a plurality of the second material texts for one first material text, and the translation unit Respectively translate the plurality of the extracted second material texts into the second translation, and the similar word extraction means includes the first translation and the plurality of translated second translations. , Each word having a predetermined similarity is extracted as a similar word.

  In this way, a plurality of second material texts for one source language document (for example, Chinese pages with the first to tenth similarities representing the high degree of similarity are extracted for a certain Japanese page) Etc.) and used to create a bilingual dictionary, by associating various words in the second language with one first language word, or conversely with many second material texts The diversity and universality of the bilingual dictionary can be improved by matching the words used.

  In another aspect (6) of the present invention, in any one of the above aspects, the material acquisition unit acquires a plurality of the first material texts, and for each of the acquired plurality of the first material texts, One or more corresponding second material texts are acquired, and for each combination of the acquired first material text and the corresponding second material text, the translation means performs the translation, and the similarity The word extraction unit extracts the similar words, the parallel translation extraction unit extracts the pair of parallel translations, and the dictionary registration unit registers in the new parallel translation dictionary.

  In this way, a plurality of first material texts are acquired, and one or more second material texts are acquired for each first material text, and the first material text and the second material text are obtained. By translating each combination, extracting similar words, and registering them in the bilingual dictionary, you can register more diverse combinations of words between the first language and the second language, creating a more practical bilingual dictionary. can get.

  According to another aspect (7) of the present invention, in any one of the above aspects, the dictionary registration unit is a set of the parallel translation words related to one word of the first language, and is different in the second language. When a plurality of words related to a word are extracted by the parallel word extraction unit, the pair of parallel words that are frequently extracted are preferentially added to the parallel dictionary.

  In this way, several expressions in the second language are extracted for the same word in the first language from one or a plurality of combinations of the first material text and the second material text. When there are multiple word set candidates, frequently used words are registered in the bilingual dictionary by preferentially registering a pair of bilingual words extracted, that is, appearing frequently, in the bilingual dictionary. A bilingual dictionary with excellent versatility and accuracy can be obtained based on more commonly used universal language expressions.

  It should be noted that a category (method for the apparatus, program for the method, etc.) different from each of the above-described modes and more specific modes described below are also included in the present invention. For different categories, “means” shall be appropriately read as “process” or “step”. Further, the order of execution of processes and steps is not limited to the above, and can be changed as appropriate or can be processed collectively.

  According to the present invention, a new bilingual dictionary can be automatically generated with appropriate accuracy on the basis of a more diverse bilingual dictionary.

The functional block diagram which shows the structure of embodiment of this invention. The figure which illustrates the information (data) used by embodiment of this invention. The flowchart which shows the process sequence in embodiment of this invention. The conceptual diagram showing the effect | action of embodiment of this invention.

  Next, modes for carrying out the present invention (referred to as “embodiments”) will be described with reference to the drawings. It should be noted that assumptions common to those already described in the background art and problems are omitted as appropriate.

[1. Constitution〕
The present embodiment relates to the bilingual dictionary generation device 1 (hereinafter abbreviated as “this device 1” or “this device”) shown in FIG. 1, and this device 1 is a first language (here, Japanese, Information processing for generating a bilingual dictionary (hereinafter referred to as a “new bilingual dictionary”) for translating a “language” (hereinafter also referred to as “original language”) into a second language (herein, Chinese, hereinafter also referred to as “target language”) Device. The apparatus 1 has a general computer configuration including at least an arithmetic control unit 6 such as a CPU, a storage device 7 such as a main memory and an auxiliary storage device, and a communication network N (Internet, mobile phone network, LAN, etc.) Communication means 8 (such as a LAN adapter or a communication gateway device).

  In the apparatus 1, a predetermined computer program (not shown) stored (installed) in advance in the storage device 7 controls the arithmetic control unit 6 so that elements (10, 20, etc.) shown in FIG. Is realized. Among these elements, the information storage means can be realized in the storage device 7 in any format such as various databases (also referred to as “DB”), variables such as files and arrays, various stacks and registers, and system setting values.

  Among such storage means, the bilingual dictionary storage means 15 is a first bilingual dictionary JE for translating the original Japanese into a third language (herein, English, also referred to as “provisional language” hereinafter). And a second bilingual dictionary CE for translating Chinese as the target language into English as the temporary language and a new bilingual dictionary JC (for example, FIG. 2 (1)) as data. is there.

  Among these bilingual dictionaries, the first bilingual dictionary JE and the second bilingual dictionary CE are stored in advance. A new bilingual dictionary JC is generated by a process described later and used by an automatic translation system (not shown). Each means other than the storage means is a processing means for realizing and executing the following information processing functions and operations.

[2. Action and effect)
FIG. 3 is a flowchart showing a processing procedure in which the apparatus 1 configured as described above generates a bilingual dictionary. An example of processing corresponding to this processing procedure is shown in the conceptual diagram of FIG.
[2-1. (Get material text)
In the processing procedure of FIG. 3, first, the material acquisition means 10 is the same as the first material text J (hereinafter referred to as “source language document” or “source language document J”) in Japanese as the source language, and this source language document J. Alternatively, the second material text C (hereinafter referred to as “target document” or “target document C”) in Chinese which is the related language and the target language is acquired (step S1).

[2-2. (Acquired by web search)
Here, the source language document J and the target document C may be material texts prepared in advance, but it is desirable to obtain them from a web search system. In this case, as shown in FIG. 4, the material acquisition unit 10 generates a source language web page (hereinafter, “source language page” or “source language page”) based on one keyword KW from a predetermined web search system W corresponding to Japanese. A first web search result P1 including a plurality of “Japanese pages” is acquired (step S11), and a predetermined web search system W (which may be the same as or different from the web search system, Chinese Based on the keyword KW, a second web search result P2 including a plurality of web pages in the target language (hereinafter referred to as “target language page”) is acquired (step S12).

  In addition, the material acquisition unit 10 includes a source language page and a target having a predetermined common characteristic among the source language page included in the first web search result P1 and the target language page included in the second web search result P2. A language page (actually, text included in a web page) is acquired as a source language document J and a target document C, respectively (step S13).

  In this way, each page selected based on the commonality among the source language page group and the target language page group searched and hit based on the same keyword in the web search system is used as the source language document and target document to be translated. Acquiring it increases the probability that similar words and corresponding words appear in similar usages and contexts in each document and translation, so the accuracy of the bilingual dictionary based on it increases.

[2-3. Examples of common features)
As described above, an example of the predetermined common feature that is a reference for acquiring the source document J and the target document C by web search is that the source language page and the target language page include the same or similar images. is there. The determination of the same similarity between images may be made by appropriately selecting a conventional technique for determining the sameness or similarity using the similarity of feature vectors representing the feature amount of each part of the image. And, for example, if there is an upper page (for example, up to 20th) source language page and target language page that include the same image in each search result, they are acquired as the source language document J and target document C, and the same If there is no such image, it may be possible to obtain the documents having the next highest similarity as the source language document J and the target document C.

  In this way, by acquiring a set of a source language document and a target document based on commonality of images, images often specify topics in a straightforward manner, so an accurate bilingual dictionary based on a set of documents related to a common topic Can be easily realized.

  Another example of the predetermined common feature is the closeness of the relative ranking of the click rate in the first web search result or the second web search result, more specifically, the first web search result. The relative rank of the click rate of the original language page in P1 and the relative rank of the click rate of the target language page in the second web search result P2 are close. For example, the relative rank is not limited to the top click rate, but may be a predetermined rank range (for example, included in the second to fifth ranks).

  Such a click rate is obtained by, for example, accumulating a web search log as illustrated in FIG. 2 (2) in the web search system W, and selecting either of the search results P1 and P2 pages based on the keyword (search keyword) KW. It is conceivable to obtain or calculate based on click information such as the number of clicks and CTR updated each time a link to a web page is clicked.

  In this way, by acquiring a pair of a source document and a target document having a common relative click rate when the web search result of the source language or target language is displayed on the access source terminal, the degree of popularity, the degree of attention, etc. An accurate bilingual dictionary can be easily created based on a set of documents with similar positioning.

[2-4. (Translation of documents and extraction of similar words)
Subsequently, the translation unit 20 translates the original language document J acquired as described above into the first translation E1 in English as a temporary language using the first parallel translation dictionary JE, and the acquired target document. C is translated into a second translation E2 in English, which is a provisional language, using the second bilingual dictionary CE (step S2). Then, the similar word extraction unit 30 extracts words having a predetermined similarity between the translated first translated sentence E1 and the second translated sentence E2 as similar words (step S3).

  For example, in the example of FIG. 4, based on the same keyword KW of “4WD”, a first search result P1 related to the Japanese website and a second search result P2 related to the Chinese website are obtained and acquired from there. Consider a case where the same word “car” is extracted as a similar word from the translated sentences E1 and E2 of the original document J and the target document C.

[2-5. (Extraction and registration of bilingual words)
Subsequently, the bilingual word extraction means 40 uses the Japanese word (for example, “car”) corresponding to the extracted similar word (here “car”) in the source language document J, and the extracted similar word (“car”). ) And a Chinese word (for example, “train”) corresponding to the target document C is extracted as a pair of parallel words (step S4). Note that the Chinese example is represented in traditional characters such as “train” due to character restrictions, but corresponds to the simplified character in the example of FIG. The dictionary registration unit 50 registers the pair of parallel translations extracted in this way ("automobile" → "train") in the new parallel translation dictionary JC (step S5).

[2-6. Basic effect)
As mentioned above, prepare the same or similar material text for both the first and second languages (here, Japanese and Chinese) for which you want to create a new bilingual dictionary, ) Are extracted from the translated sentences translated into (), and Japanese and Chinese words corresponding to the extracted words are associated to form a bilingual dictionary. Thus, if there is a bilingual dictionary of Japanese → English and Chinese → English with appropriate accuracy, even if there is no English → Chinese bilingual dictionary or the accuracy is inferior, based on the similarity between words translated into English, A bilingual dictionary between Japanese and Chinese can be automatically created with appropriate accuracy.

[2-7. Example using multiple target documents)
In addition, parallel translation words may be extracted from the source language document J using a plurality of target documents C. In this case, the material acquisition unit 10 acquires a plurality of target documents C for one source language document J, and the translation unit 20 translates the extracted plurality of target documents C into the second translated sentence E2, respectively. The similar word extraction unit 30 extracts words having a predetermined similarity between the first translated sentence E1 and the plurality of translated second translated sentences E2 as similar words.

  As described above, a plurality of target documents C (for example, Chinese pages having the first to the tenth similarities representing the high degree of similarity with respect to a certain Japanese page are extracted for one source language document J. Etc.) and used for the creation of a bilingual dictionary, various words of the target language can be associated with one source word (for example, “car” → “train”, “car” → “ "Carriage"), and conversely, words used in many target documents can be matched to improve the diversity and universality of the bilingual dictionary.

[2-8. Example of using multiple source language documents)
Although an example using a plurality of target documents C has been shown above, a plurality of source language documents J may be used. In this case, the material acquisition unit 10 acquires a plurality of source language documents J, acquires one or more corresponding target documents C for each of the acquired source language documents J, and the source language documents thus acquired. For each combination of target document C corresponding to J, translation means 20 translates, similar word extraction means 30 extracts similar words, parallel translation word extraction means 40 extracts a pair of parallel translation words, and registers the dictionary. The means 50 registers in a new bilingual dictionary.

  In this way, a plurality of source language documents J are acquired, one or more target documents C are acquired for each source language document J, translation is performed for each combination of source language document J and target document C, and similar words are extracted. By registering in the bilingual dictionary, more various combinations of words between the original language and the target language can be registered, and a more practical bilingual dictionary can be obtained.

[2-9. (Registration based on the number of extractions)
In addition, when there are a plurality of candidate words of the target language with respect to one word of the original word, it is desirable to prioritize the one with a large number of extractions. In this case, the dictionary registration means 50 is a case where a plurality of parallel word extraction means 40 that are pairs of parallel words related to one word of the original language and that are different from the target language are extracted (for example, “car” → ("Train" and "car"->"car")), the pair of translated words with a high number of extractions will be given priority and added to the new bilingual dictionary.

  In this way, when the source language document J and the target document C are combined in one or more combinations, the expression in the target language is extracted for the same word in the source language, and there are a plurality of pairs of parallel word combinations. By registering a pair of translated words with a high number of times of extraction or appearance in the bilingual dictionary, frequently used words are preferentially registered in the bilingual dictionary. Based on the above, a bilingual dictionary with excellent versatility and accuracy can be obtained.

[2-10. (Judgment of translation accuracy)
In the present invention, there are only bilingual dictionaries other than Japanese → English, and there are only middle → English, or middle → English and English → middle, but English → medium is more accurate than middle → English. Intended for cases. Such translation accuracy can be measured by, for example, translating a certain page or string of words by using a translation system using bilingual dictionaries of English-> Medium and Middle-> English, and comparing the correct answer rates.

  It is conceivable to use BLUE (Non-Patent Document 1), NIST, WER (Word Error Rate) or the like as an index indicating the accuracy of such a translation result. These are quantified by comparing the translation result with the model answer. For example, if there are Japanese → English as well as Japanese → English as a bilingual dictionary, middle → English and English → middle, English → middle rather than English → middle If it is highly accurate, create a new bilingual dictionary for Japanese → English by combining Japanese → English and English → Chinese. If English → Chinese → English → Middle is more accurate, as explained above It is conceivable to create a new bilingual dictionary for Japanese → Chinese by combining Japanese → English and Chinese → English.

[3. Other embodiments]
In addition, said each embodiment is only an illustration, and this invention includes what is illustrated below and other embodiment other than that. For example, the present invention can be applied not only to bilingual dictionaries of Japanese and Chinese, Japanese and other languages (Korean, French, etc.), but also to bilingual dictionaries of other languages. In addition, the “word” related to the bilingual translation in the present application is not limited to a word or a single phrase, and may be an arbitrary unit such as a fixed phrase composed of a plurality of words.

  In addition, each element such as means relating to the present invention may be realized by other information processing mechanisms such as a physical electronic circuit without being limited to the arithmetic control unit of the computer. Further, each configuration diagram, data diagram, flowchart diagram, and the like are merely examples, and the presence / absence of each element, the order of arrangement and processing execution, specific contents, and the like can be changed as appropriate. For example, the apparatus 1 may be configured integrally with the web search system W, or conversely, may be realized by using a plurality of apparatuses such as servers. A configuration in which each storage means is realized by a separate and independent server device or system is also common. Depending on the function, the configuration can be flexibly changed, for example, by calling an external platform or the like with an API (application program interface) or network computing (so-called cloud or the like).

1 Bilingual dictionary generation device (this device)
W Web search system 6 Arithmetic control unit 7 Storage device 8 Communication means 10 Material acquisition means 15 Bilingual dictionary storage means 20 Translation means 30 Similar word extraction means 40 Bilingual word extraction means 50 Dictionary registration means KW Keyword J Original language document C Target document E1, E2 Translation JE First Bilingual Dictionary CE Second Bilingual Dictionary JC New Bilingual Dictionary N Communication Network

Claims (9)

In an information processing apparatus that generates a new bilingual dictionary for translating a first language into a second language based on the bilingual dictionary according to the first language and the second language,
A first bilingual dictionary for translating the first language into a third language; a second bilingual dictionary for translating the second language into the third language; and the new bilingual dictionary And storage means for,
Material acquisition means for acquiring a first material text in the first language and a second material text in the second language that is the same as or related to the first material text;
The obtained first material text is translated into a first translation in the third language using the first bilingual dictionary, and the obtained second material text is translated into the second bilingual dictionary. Translation means for translating into a second translation in the third language,
Similar word extraction means for extracting a word having a predetermined similarity between the translated first translated sentence and the second translated sentence as a similar word;
A word in the first language corresponding to the extracted similar word in the first material text; a word in the second language corresponding to the extracted similar word in the second material text; Bilingual word extraction means for extracting
Dictionary registration means for registering the extracted pair of parallel translation words in the new parallel dictionary;
An information processing apparatus comprising: The material acquisition means includes
Obtaining a first web search result including a plurality of web pages in the first language based on one keyword from a predetermined web search system;
Obtaining a second web search result including a plurality of web pages in the second language based on the keyword from a predetermined web search system;
Of the web page in the first language and the web page in the second language, the web page in the first language and the web page in the second language having predetermined common features are respectively connected to the first language web page. The information processing apparatus according to claim 1, wherein the information processing apparatus acquires the material text and the second material text. 3. The predetermined common feature is that the web page in the first language and the web page in the second language include images that are the same or similar to each other. Information processing device. The information processing apparatus according to claim 2, wherein the predetermined common feature is a closeness of a relative rank of a click rate in the first web search result or the second web search result. The material acquisition means acquires a plurality of the second material texts for one first material text,
The translation means translates the plurality of the extracted second material texts into the second translated text,
The similar word extracting unit extracts words having a predetermined similarity between the first translated sentence and the plurality of translated second translated sentences as similar words, respectively. The information processing apparatus according to any one of 1 to 4. The material acquisition means acquires a plurality of the first material texts, acquires one or more corresponding second material texts for each of the plurality of acquired first material texts,
For each combination of the second material text corresponding to the acquired first material text, the translation unit performs the translation, the similar word extraction unit extracts the similar word, and the bilingual word extraction The information processing apparatus according to any one of claims 1 to 5, wherein means extracts the pair of translated words and the dictionary registration means registers in the new parallel dictionary. The dictionary registration means
In the case where a plurality of bilingual word extraction means that extract a pair of parallel words related to one word of the first language and related to different words of the second language, the number of times of extraction is large. The information processing apparatus according to any one of claims 1 to 6, wherein a pair of parallel translation words is preferentially added to the parallel translation dictionary. In an information processing method for generating a new bilingual dictionary for translating a first language into a second language using a computer based on the bilingual dictionary according to the first language and the second language,
A first bilingual dictionary for translating the first language into a third language; a second bilingual dictionary for translating the second language into the third language; A bilingual dictionary and storage means for
A material acquisition process in which the computer acquires a first material text in the first language and a second material text in the second language that is the same as or related to the first material text. When,
The computer translates the acquired first material text into the first translation in the third language using the first bilingual dictionary, and converts the acquired second material text into the second translation text. A translation process for translating into a second translation in the third language using a bilingual dictionary;
A similar word extraction process in which the computer extracts a word having a predetermined similarity between the translated first translated sentence and the second translated sentence as a similar word;
A computer in which the first language word corresponds to the extracted similar word in the first material text, and the second language word corresponds to the extracted similar word in the second material text. Bilingual word extraction processing for extracting a word as a pair of bilingual words;
A dictionary registration process in which the computer registers the extracted pair of bilingual words in the new bilingual dictionary;
An information processing method comprising: An information processing program for generating a new bilingual dictionary for translating a first language into a second language based on the bilingual dictionary according to the first language and the second language by controlling a computer And the program is
A first bilingual dictionary for translating the first language into a third language; a second bilingual dictionary for translating the second language into the third language; A bilingual dictionary and storage means for
Causing the computer to obtain a first material text in the first language and a second material text in the second language that is the same as or related to the first material text;
The computer causes the acquired first material text to be translated into a first translation in the third language using the first bilingual dictionary, and the acquired second material text is converted to the second Translated into a second translation in the third language using the bilingual dictionary of
Causing the computer to extract words having a predetermined similarity between the translated first translated sentence and the second translated sentence as similar words;
The computer further includes the first language word corresponding to the extracted similar word in the first material text, and the second language word corresponding to the extracted similar word in the second material text. Words and words as a pair of translated words,
An information processing program for causing a computer to register the extracted pair of translated words in the new parallel dictionary.

Images