JP2004185641A - Device for retrieving example sentence in two languages - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To prepare the translated sentence of an object language by only inputting the word or the idiom of a source language. <P>SOLUTION: A device for retrieving the example sentence in two languages comprises: inputting the document of the source language and the translated sentence document of the object language by a two language input means; storing the sentence of the source language, the sentences of a character string and the translated sentence document, and the character string on a two language database by utilizing corresponding information by a translated sentence database preparation means; then, extracting the candidate of a word used in a translation and complying with retrieval request for the character string of the source language by a two language database retrieval means; changing the candidate of the word used in the translation to a word used in a translation by a template change means to apply the word used in the translation to the template; and displaying the translated sentence of the object language by a display means. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

  The present invention uses a bilingual database in which sentences and words and idioms of a bilingual document are associated with each other, in order to know how a certain word or phrase in the native language is expressed in the target language, The present invention relates to a bilingual example sentence search device that searches for a translation example of a target language in a bilingual document, and supports creation of a document in the target language based on the retrieved example.

  There are more opportunities for Japanese to write business documents and dissertations in English. However, when a person with poor English skills prepares an English document, it tends to be a poor English sentence. For this reason, publications such as a bilingual example sentence collection of Japanese and English expressions and a fixed phrase example collection, an electronic dictionary with example sentence search, a word processor attached to a fixed document script collection, and the like have become widespread.

  However, in general, an appropriate expression desired by an English document creator is hard to find in an existing example sentence collection, and is often not an expected English translation. As a result, the current situation is that after all, the format and expression of the teachings will be reluctant and the English expressions will be poor.

  To deal with such a problem, for example, it is shown in “Translational alignment using dictionaries and statistics, Masahiko Haruno, Takefumi Yamazaki (NTT), Proc. There was a way. This method is a method of semi-automatically associating a sentence unit of an existing bilingual document using knowledge of the bilingual dictionary and statistical information obtained from the document, and also presents a result of the association to a user. Is disclosed.

  When a large number of existing bilingual documents that the user wants to use as a bilingual example sentence are given to this system, a bilingual database with sentence correspondence can be semi-automatically created. In this way, if a user creates a bilingual database of bilingual documents that the user wants to use as an example of English creation, if he wants to know the English translation of a certain Japanese expression, he can search the bilingual database with that Japanese expression and find the Japanese translation. A bilingual sentence including a word expression can be extracted, and an English sentence can be created based on the sentence.

  However, in the above-mentioned conventional method, since only the correspondence of the sentence level is attained, it is not possible to know even English translation words such as words and idioms. In other words, even if the English text and the Japanese text in the bilingual document database are correlated in units of one sentence, and the user can search for a bilingual sentence that includes the specific expression desired by the user, even if the specific expression corresponds to the corresponding translated sentence. It is not easy for the user to find out where to respond.

  Further, since the bilingual document is stored in the original sentence state, for example, a search key of “input” cannot extract an example of “input and to”. That is, it is not possible to search for a word such as a verb whose inflection changes in a standard form. On the other hand, translation dictionaries with an example search function that are already commercially available on media such as CD-ROMs have a search function in a standard form, but bilingual example sentences and search keys there are special expressions in a specific field. Is limited to

  From such a point, a bilingual database creation device capable of creating a bilingual database having information on correspondence between words and idioms as well as sentences, and a purpose for an arbitrary character string in a source language using the bilingual database. It has been desired to realize a bilingual example sentence retrieval device capable of extracting a translation of a language. Further, it has been desired to realize a bilingual example sentence search device for supporting creation of a target language document.

The present invention employs the following configuration to solve the above-described problem.
<Structure of Claim 1>
A bilingual document input means for inputting a target language document to be used as a template serving as a template for applying the source language expression, and a source language based on the bilingual document between the source language and the target language input by the bilingual document input means. And a corresponding translation information of the sentence to the target language, a bilingual database creating means for creating a bilingual database having the corresponding information of the character strings in each sentence, receiving a search request for any character string in the source language, from the bilingual database A bilingual database search means for extracting a translation target candidate of a corresponding target language, and a template for changing a source language change requested expression to a target language translation based on the extraction result of the bilingual database search means and applying the target language to the template Bilingual example sentence detection, comprising: a change unit; and a display unit for displaying a sentence in the target language changed by the template changing unit. It is a device.

<Explanation of Claim 1>
The source language and the target language are, for example, Japanese and English, but other languages can be similarly applied. The character string in the source language is, for example, a word or idiom, but is not limited to a word or idiom as long as the character string associates the source language with the target language in one sentence.

  The user inputs the target language document to be used as an example together with the source language document by the bilingual document input means. Sentences are correlated by dividing each document into one sentence, performing morphological analysis, and calculating the similarity between sentences in each language using a word dictionary. By selecting such a combination, the sentence association process is performed. This process may be performed by other methods.

  String matching involves finding pairs of source language words and target language words that appear frequently and frequently in source language language and target language language sentences that are associated on a sentence basis, based on the associated sentences. Is used to associate words and idioms. Also, this processing may use another method.

  The bilingual database creation means creates a bilingual database having correspondence information of corresponding sentences and correspondence information of character strings. That is, for example, by simply inputting a desired bilingual document such as a bilingual document in a specialty field, in that field, the expression in the source language corresponds to any translated word, and what is the sentence in that case Create a database in which sentences, words, and idioms are associated with each other.

  In the present invention, an arbitrary expression in a source language sentence is designated using the bilingual database thus created as a template, thereby creating a sentence in a target language corresponding to the designation.

  The bilingual database registers a bilingual document that the user wants to use as a template. In this case, it is desirable to register as many types of bilingual documents as possible. When the user selects a bilingual document to be used as a template and instructs to change the expression of an arbitrary part, the template changing means sends a search request for the specified expression to the bilingual database searching means. The bilingual database search means extracts a target language corresponding to the specified expression from the bilingual database and passes it to the template changing means. The template changing means creates a sentence in which the translation of the received target language is applied to the template, and the template display means displays the sentence.

  Thus, any type of template can be prepared according to the user's preference. Further, since the change of the expression is specified in the source language, knowledge of the target language is not particularly required, and an accurate target language change sentence can be easily obtained.

<Structure of Claim 2>
2. The bilingual example sentence search device according to claim 1, wherein, as a result of the search by the bilingual database search means, when there are a plurality of translation word candidates, the independent word of each translated sentence including the translation word candidate and the independent word of the original sentence including the search key are obtained. Is a bilingual example sentence search device, characterized by comprising: a template changing unit that preferentially changes a translated sentence having the largest number of overlapping words to a translated word of a target language.

<Explanation of Claim 2>
According to a second aspect of the present invention, in the first aspect of the present invention, when there are a plurality of translation word candidates, a translation word candidate having a large number of the same words in the independent sentence of the translation including the translation word candidate and the original sentence is determined. The priority is selected. In other words, this is because a word with more duplicated independent words has a higher degree of approximation to the original sentence and is considered to be a translated word with higher accuracy.
As a result, an optimal translation of the target language is selected, and a document having a uniform expression of the target language can be obtained.

  According to the present invention, a corresponding target language document can be created only by modifying a source language document by using a template.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
"Concrete example"
<Constitution>
FIG. 1 is a configuration diagram showing a specific example of the bilingual example sentence search device of the present invention.
This apparatus stores an input / output unit 41 for inputting and outputting data such as a user inputting a bilingual document, a bilingual database creating unit 42 for associating a sentence, a phrase, and a word unit, and the bilingual database. And a bilingual database search means 43 for searching the bilingual database 44 according to the search key input by the user. In this specific example, the source language is Japanese and the target language is English.

  The input / output unit 41 includes a bilingual document input unit 411 for the user to input a bilingual document as a source of a bilingual database, a template selecting unit 412, a template displaying unit 413, and a template changing unit 414. .

  The bilingual database creation unit 42 includes a sentence association unit 421 that associates a sentence unit by a method described later, and a word / idiom association unit 422 that associates a word or idiom by a method described later. And a bilingual database processing means 423 for shaping the bilingual database with a predetermined definition based on the result of the attachment. In this specific example, a bilingual database input device 411, a sentence associating device 421, a word / idiom associating device 422, a bilingual database processing device 423, and a bilingual database 44 constitute a bilingual database creating device.

<motion>
This device is
(1) A bilingual document that the user wants to use as an example is registered as a bilingual database.
(2) Search the registered bilingual database. Each is described below.

FIG. 2 is a flowchart showing a flow until the present apparatus creates a bilingual database from a bilingual document given by a user.
FIG. 6 is an example of a bilingual document to be created in the bilingual database of the present apparatus.
The operation here will be described by taking as an example a case where the user inputs the bilingual document shown in FIG.

  The user inputs a bilingual document that he / she wants to use as a sample for creating an English sentence using the bilingual document input unit 411 (step S201). As a result, the input Japanese and English bilingual documents are passed to the sentence association unit 421. The sentence association unit 421 automatically determines which sentence in the Japanese document corresponds to which sentence in the English document (step S202). Here, for example, as shown in Japanese Patent Application No. 7-194134 filed earlier, the sentence association unit 421 divides each document into individual sentences, performs morphological analysis, and compares the Japanese sentence with a Japanese sentence using a word dictionary. The similarity of the English sentences is calculated, and association is performed by selecting a combination that maximizes the sum of the similarities in the entire document.

  Next, the Japanese document and the English document with the sentence correspondence are passed to the word / idiom association means 422. The word / idiom association unit 422 automatically determines which expression in the English document corresponds to the word or idiom appearing in the Japanese document (step S203). Here, as shown in Japanese Patent Application No. 8-178595 filed earlier, the word / idiom associating means 422 simultaneously and frequently appears in Japanese-English bilingual sentences correlated in sentence units. The correspondence between words and idioms is performed by obtaining a pair of a word and an English word.

FIG. 4 shows an example of the result of the word / idiom association means 422 in the bilingual document of FIG.
The result of the sentence associating means and the result of the word / idiom associating means are passed to the bilingual database processing means 423, and the bilingual database processing means 423 shapes these results into a bilingual database format (step S204).
The bilingual database formatted according to this definition is stored in the bilingual database 44 (step S205), and the bilingual database creation processing ends.

FIG. 7 shows an example of the definition of the bilingual database written in BNF (Backus-Naur form).
8 and 9 show the bilingual database of the bilingual document of FIG. 6 which is shaped based on the definition.

  The bilingual database 44 stores two types of forms, an appearance form that holds the original text of the input bilingual document and a standard form (see 61 in FIG. 7). The document association information is the document identification number in {English document tag} and {Japanese document tag} (see 62a and 62b), and the sentence association information is the sentence identification number in {sentence tag}. (See 63), and information on the correspondence between words and idioms is stored in the word identification number in {word tag} (see 64), respectively.

  A specific example will be described with reference to the bilingual database shown in FIGS. The document identification number of this bilingual document is ID = 1 (71a, 71b) by {English document tag} and {Japanese document tag}. In addition, the English sentence “typing n or RETURN skips them.” And the Japanese sentence “If you enter n or RETURN, the instructions are skipped.” Are supported by SENT ID = 2 (75a, 75b). It is attached. Further, "answering" in the English sentence with ID = 1 and "input" in the Japanese sentence are associated with R ID = 3 (73a, 73b). In the case of an idiom, one word identification number is assigned to a plurality of words such that R ID = 2 (74a, 74b) is assigned to both words “start” and “up”. I do.

  As described above, in this specific example, since the bilingual database 14 can be created not only in units of sentences but also in units of words and idioms, the user only needs to input a bilingual document in a desired field. In that field, it is possible to obtain a database having correspondence information such as what translation of the Japanese expression corresponds to English and what the sentence in that case is.

Next, a method of searching for a bilingual example sentence from the created bilingual database 44, which is the operation (2), will be described.
FIG. 3 shows the flow of this processing.
Here, a case will be described in which the user searches the bilingual database registered above for a bilingual example sentence related to “input”.

  The user inputs “input” using a search key input unit that is not disclosed (step S301). Thereby, the bilingual database search means 43 receives the search key "input" and searches the range of {Japanese document} (77 in FIG. 8) in the bilingual database 44 (step S302). First, search by appearance.

  The search method in the appearance form is as follows. Take out the appearances of the Japanese sentence of <SENT ID = 1> in Fig. 8 and connect them, as shown in "When starting addbib, the instruction is given by typing y at the first Instructions? Prompt." The original text can be restored. In the restored sentence, "input" is searched, and if a search key exists, {Japanese document tag}, {sentence tag} containing the word, and {word tag attached to the word} Extract｝. Here, <TEXT LANG = JAP ID = 1> (71a), <SENT ID = 1> (72a), and <R ID = 3> (73a) are extracted. The above search is repeated for all the sentences in the bilingual database, and if there is no more sentence to be searched, the process in the appearance form ends (step S303).

Next, in the standard form, a search process is performed in the same manner as in the appearance form (step S304).
In the example of the Japanese search key "enter", <TEXT LANG = JAP ID = 1> (71a), <SENT ID = 2> (75a), and <R ID = 1> (76a) are extracted.
When there is no more sentence to be searched (step S305), it is determined whether or not a search result exists (step S306). In step S306, if there is a search result, from the range 78 of {English document} of the bilingual database 44, the extracted English document tag, sentence tag, and sentence tag having the same identification number as the identification number of each of the extracted Japanese document tag, sentence tag, and word tag. The English sentences 72b and 75b having the tag and the English words 73b and 76b having the identification number of the same word tag are extracted (step S307), and the English sentence and the English word are displayed together with the Japanese sentence using the search result display means 113. (Step S308).
On the other hand, if the Japanese search key does not exist in both the standard form and the appearance form, the absence of the Japanese search key is displayed using the search result display means (step S309).

FIG. 10 is an example of the output result.
As shown in the figure, a bilingual example sentence including the Japanese search key "enter" given by the user is displayed, and the English words ("answer" and "type") corresponding to "enter" are included in the bilingual example sentence. Is highlighted.

  The English translation of a Japanese word is often not uniquely determined, and multiple candidates are possible. For this reason, it is difficult to determine which English word should be used in an English document to be created. In such a case, if you have an English document and its Japanese translation that you want to use as a sample of English text creation, register the bilingual document in this device as a bilingual database, and search the database in Japanese. An English expression corresponding to the Japanese expression can be easily extracted, and an English sentence can be created based on the result.

  For example, suppose that the user wants to create an English sentence that means "input" with the same content as a bilingual document registered in the apparatus. By performing a search using this apparatus, it is found that the English translation of “input” is either “answer” or “type”. By referring to the bilingual example sentence output at the same time, it is possible to determine whether “answer” or “type” is appropriate, and create an English sentence.

  For example, if "input" is searched in a general bilingual dictionary, only the English translation of "input" can be obtained. However, in an actual document, various English is used depending on the context of the document, such as “answer” and “type”. If various bilingual documents are registered, the present apparatus can also be used as a translation dictionary with bilingual example sentences according to actual documents.

  In addition, since the Japanese key for searching can be searched not only in a form appearing in a bilingual document but also in a standard form, a more accurate translated word can be extracted.

  The apparatus further includes a bilingual dictionary 45 in which bilingual translations of general words are stored, a bilingual document input unit 411 for the user to input a bilingual document from which a bilingual database is created, and A template selection unit 412 for selecting a bilingual document, a template display unit 413 for displaying a bilingual document specified by a user as a template, and a user or bilingual database search unit for changing the displayed template. It has a template changing means 414 for performing.

<motion>
In this apparatus, a user selects a bilingual document (template) from the bilingual database and changes the Japanese document of the bilingual document to create an English document.
Here, it is desirable to register as many types of bilingual documents as possible.

Next, an operation using a template will be described.
FIG. 5 shows the flow of processing until the user creates an English document by using a bilingual document in the bilingual database 44 in the apparatus as a template.
The user selects a bilingual document to be used as a template from the bilingual database 44 by the template selecting means 412 (step S501). As a result, the bilingual database search unit 43 extracts the bilingual document from the bilingual database 44, and the template display unit 413 displays it (step S502).
FIG. 11 and FIG. 12 are explanatory diagrams showing this display example.

  The user changes the Japanese document of the template (step S503). For example, a case will be described in which the dotted line portion “Japanese tea” of a Japanese document is changed to “scarf” and “trees and flowers” is changed to “landscape”. The template changing unit 414 extracts the word tag assigned to the changed part from the bilingual database 44 via the bilingual database searching unit 43.

FIG. 13 shows a part of the bilingual database in the templates of FIGS.
In the example of "Japanese tea", <TEXT LANG = JAP ID = 20> (1201a), <SENT ID = 5> (1202a), and <R ID = 4> (1203a) are extracted.

  Next, the bilingual database search means 43 searches the bilingual database 44 using the changed word Japanese as a search key, and extracts the corresponding English translated word (step S504). The extraction method is in accordance with the first embodiment. In the above example, “scarf” and “landscape” are search keys.

  As a result of the search in step S504, if the search is successful (step S505) and the English expression corresponding to the search key is uniquely determined, the English translation is the same as the original Japanese expression sentence tag and word tag. The position of the English document having the identification number is replaced with English (step S508). For example, if the English translation of “scarf” is uniquely determined as “silky scarf” as a result of the search, <TEXT LANG = ENG ID = 20> (1201b), <SENT ID = 5> (1202b), < “Japanese＼＼Japanese tea＼＼tea” of R ID = 4> (1203b) is changed to “silky＼＼silky scarf＼＼scarf”.

On the other hand, if a plurality of English translation word candidates exist as a result of the search, one of the most suitable English translation words is selected and changed. An example of the selection method will be described using “landscape” as an example. As an example sentence with "landscape",
a. Scenery seen from the train window
A scene from the train window.
b. “The Valley” is recommended as a beautiful landscape book
I can recommend “The Valley” as a book with fine landscape paintings.
If two are found, two English translations of "scene" and "landscape" are candidates for "landscape".

  As described above, when there are a plurality of translation word candidates, the independent word constituting the sentence and the English translation word of the sentence having the larger number of overlapping independent words of the original sentence are prioritized. For example, a. Is {watch the train window}, b. Is composed of the independence of “The Valley”, a beautiful landscape painting book. On the other hand, the original text is composed of the independence of "I will send a book full of photos next time". In this case, a. Does not have duplicate words, but b. Then "books" overlap. Therefore, b. The English translation of "landscape" is adopted. If no priority is given, both English translation words are displayed on the template display means 413, and the user selects one of them.

  If there is no candidate as a result of the search, the bilingual dictionary 45 is searched (step S506), and the English translation is stored in the position of the English document having the same identification number as the original Japanese expression sentence tag and word tag. (Step S508). The changed English document is displayed by the template display means (step S509). Then, after the above processing is repeated for all the parts changed by the user (step S510), the processing ends.

14 and 15 are explanatory diagrams illustrating examples of output results.
The portion of the Japanese document that has been changed by the user is shown enclosed by squares, and the change is also reflected in the English document.

<effect>
In this specific example, since the user can create a template by himself / herself, any type of template can be prepared according to the user's preference.

  In addition, when an English document is created using a template, a method of modifying a portion that needs to be changed on the English document template has conventionally been adopted. Therefore, the user needs the ability to understand the English document as a template. However, if this apparatus is used, an English document can be created only by correcting a Japanese document having the same content as the English document, and therefore, knowledge of English is not required.

  Further, in this specific example, the Japanese English translation word used when changing the template is determined by searching the bilingual database 44, and if there are a plurality of translation word candidates, the constituent words of the sentence are inspected. Therefore, the English document created by the present apparatus is a document having a uniform English expression.

1 is a configuration diagram of a bilingual example sentence search device of the present invention. It is a flowchart of a bilingual database creation of the bilingual example sentence search device of the present invention. It is a flowchart which shows the flow of a bilingual example sentence search of the bilingual example sentence search device of this invention. FIG. 9 is an explanatory diagram showing a result of a word / idiom association unit. 5 is a flowchart illustrating an operation of using a template in the bilingual example sentence search device of the present invention. FIG. 5 is an explanatory diagram showing an example of a bilingual document to be created in a bilingual database of the bilingual example sentence search device of the present invention. It is explanatory drawing which shows the example of a definition of a bilingual database. It is explanatory drawing (the 1) which shows the example of a bilingual database. FIG. 9 is an explanatory diagram (part 2) illustrating an example of a bilingual database. FIG. 9 is an explanatory diagram showing an example of an output result of a bilingual example sentence search in the bilingual example sentence search device of the present invention. It is explanatory drawing (the 1) which shows the example of the template in the bilingual example sentence search device of this invention. It is explanatory drawing (the 2) which shows the example of the template in the bilingual example sentence search device of this invention. It is explanatory drawing which shows the example of the bilingual database of the bilingual example sentence search device of this invention. FIG. 9 is an explanatory diagram (part 1) illustrating an example of an output result of the bilingual example sentence search device of the present invention. It is explanatory drawing (the 2) which shows the example of the output result of the bilingual example sentence search device of this invention.

Explanation of reference numerals

43 bilingual database searching means 44 bilingual database 411 bilingual document input means 421 sentence associating means 422 word / idiom associating means 423 bilingual database processing means 413 template displaying means 414 template changing means

Claims (2)

A bilingual document input means for inputting a target language document to be used as a template for applying a source language expression,
Based on the bilingual document between the source language and the target language input by the bilingual document input means, a bilingual database for creating a bilingual database having correspondence information of sentences for the source language and the target language and correspondence information of character strings in each sentence. Database creation means;
A bilingual database search means for receiving a search request for an arbitrary character string in the source language and extracting a corresponding word candidate in the target language from the bilingual database;
Based on the extraction result of the bilingual database search means, a template change means for changing the expression requested to be changed in the source language to a translated word in the target language and applying it to the template,
A display unit for displaying a sentence in the target language changed by the template changing unit. 2. The bilingual example sentence search device according to claim 1,
If there are multiple candidate translations as a result of the search by the bilingual database search means, priority is given to the translation with the largest number of independent words in each of the translations containing the translation candidate and the original sentence including the search key. And a template changing means for changing to a translation of the target language.

Images