JP2003196309A - Document retrieval device, document retrieval method and program for making computer perform the same method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To retrieve a document in an optional language from an optional language by a simple processing and with high accuracy. <P>SOLUTION: The document is retrieved on the premise that a Japanese document and an English document as a part of a group of document to be retrieved are in faithful relation of parallel translation (the description contents are identical and only languages describing the contents are different). First, by using a natural sentence inputted in Japanese as retrieval conditions, a Japanese sentence conforming to the conditions is retrieved (shown in Fig. (1)). Then, when documents J-1 and J-2 are defined to conform to the conditions, since a document E-1 in the relation of parallel translation exists in the document J-1 among the conforming documents, by using a natural sentence in English as a text of the English sentence E-1 as new retrieval conditions the English sentence conforming to the conditions is retrieved next (shown in Fig. (2)). Then, a document conforming to the new conditions, namely, the document E-1 itself, documents E-2 and E-3 similar to the document E-1 are defined as the last retrieval results. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document retrieval apparatus, a document retrieval method and a program for causing a computer to execute an electronic document described in a language different from a retrieval condition described in a language. Regarding

[0002]

2. Description of the Related Art A document search in which a language of a document to be searched and a language of a search condition are generally called "multilingual document search", for example, a document written in English is searched by a Japanese keyword. Was possible from.

As an example, Cyberspace Laboratories is named “TITAN”, and AltaVista Company is named “AltaVista”, and the language is W in each language.
The EB document search service has already been put to practical use. As a research paper, "Multilingual Internet Information Retrieval Joint Research Project in AMF" (NTT
NEWS RELEASE 1999/02/24).

[0004]

However, in any of the above-mentioned conventional techniques, the search condition is converted into the language of the search target document, or conversely, the search target document is converted into the language of the search condition. The search was performed after standardizing the language of.

For this conversion, a translation system between a plurality of languages and at least a language replacement system at the word level are required, which requires a large amount of cost for preparing and refining data such as a translation dictionary, and at the time of executing a search. The load on the system will also increase. Moreover, since ambiguous words often cause deviations in meaning or concept during the translation process, noise that is not intended by the operator is likely to be included in the search results, and the search accuracy is commensurate with the complexity and complexity of processing. There was a problem that was not obtained.

Attempts have been made to statistically process a large number of documents to identify correspondences between words in a plurality of languages without using a dictionary (Japanese Patent Laid-Open No. 2001-4323).
6) It requires complicated and large-scale processing, and at the current technical level, it cannot be expected to be as accurate as a dictionary created by hand.

The present invention has been made in view of the above-mentioned problems of the prior art, and is a document search capable of accurately searching a document of an arbitrary language from an arbitrary language with a simple process. An object is to provide an apparatus, a document search method, and a program that causes a computer to execute the method.

[0008]

[Means for Solving the Problems]
In order to achieve the object, the document search device according to the invention described in claim 1 is a document search device for searching an electronic document described in a second language from a search condition described in a first language, First search means for searching an electronic document described in the first language for electronic documents that match the search condition described in the first language; and an electronic document searched by the first search means. Second search means for searching an electronic document which is a parallel translation of the document from the electronic document described in the second language, and the first document based on the text of the electronic document searched by the second search condition. Generating means for generating a search condition described in the second language, and searching for an electronic document matching the search condition generated by the generating means from the electronic document described in the second language. Characterized by comprising a search unit of the.

According to the first aspect of the present invention, a complex work (for example, machine translation) for eliminating the language difference between the search condition and the document to be searched while realizing the multilingual document search. Is not done at all.

According to a second aspect of the present invention, there is provided a document search device according to the first aspect, wherein the first search means and / or the third search means are:
It is characterized in that an electronic document whose text is semantically similar to a natural sentence constituting the search condition among the electronic documents is searched as an electronic document that matches the search condition.

According to the second aspect of the present invention, only the documents such as fields and topics whose main meaning of the entire text matches the search condition are searched.

According to a third aspect of the present invention, there is provided the document search device according to the first or second aspect of the invention, wherein the second search means is searched by the first search means. It is characterized in that, for all the electronic documents described above, a parallel translation of the electronic document is searched from the electronic documents described in the second language.

According to the invention of claim 3, the second
The search condition in the language is generated by using the bilingual translation of the document obtained in the search in the first language without omission.

According to a fourth aspect of the present invention, there is provided the document search device according to the first or second aspect of the invention, wherein the second search means is searched by the first search means. The electronic document which is a parallel translation of a part of the electronic documents among the electronic documents is searched from the electronic documents described in the second language.

According to the invention described in claim 4,
The search condition in the language is generated by selectively using only the bilingual translations of the documents obtained by the search in the first language that have a particularly high degree of matching with the search condition.

A document search device according to a fifth aspect of the present invention is the document search device according to any one of the first to fourth aspects of the present invention, further including the first search means. Information that can identify electronic documents and the third
It is characterized by further comprising display means for displaying information capable of specifying the electronic document retrieved by the retrieval means.

According to the invention of claim 5, the first aspect
The result of the search in the language and the result of the search in the second language are displayed together on the screen.

Further, a document search method according to a sixth aspect of the present invention is the document search method for searching an electronic document described in a second language from a search condition described in a first language. A first search step of searching an electronic document described in the first language for an electronic document that matches the search condition described in the first language; and an electronic document searched in the first search step. A second search step of searching for an electronic document which is a parallel translation of the electronic document described in the second language, and the second based on the text of the electronic document searched by the second search condition. A third step of searching the electronic document described in the second language for an electronic document matching the search condition generated in the generating step, and generating a search condition described in the second language. Work And wherein the containing when the.

According to the invention described in claim 6, a complicated work (for example, machine translation) for eliminating the language difference between the search condition and the document to be searched while realizing the multilingual document search. Is not done at all.

A program according to a seventh aspect of the invention causes a computer to execute the method described in the sixth aspect.

According to the invention described in claim 7, the method described in claim 6 is executed by a computer.

[0022]

BEST MODE FOR CARRYING OUT THE INVENTION Preferred embodiments of a document search device, a document search method and a program for causing a computer to execute the method according to the present invention will be described in detail below with reference to the accompanying drawings.

(Basic Principle of the Invention) Before starting the description of a specific embodiment, the basic principle of the present invention will be briefly described. FIG. 1 is an explanatory view schematically showing the basic principle of the present invention. The search target document group in the present invention is assumed to be a large number of WEB documents collected from the Internet, for example. In the figure, a document with an ID starting with J indicates a document written in Japanese, and a document with an ID starting with E indicates a document written in English.

Here, the websites of companies that operate on a global scale or companies that open stores on the e-marketplace may be created in multiple language versions such as Japanese and English. In the figure, document J-1 shows a Japanese version of such a home page, and document E-1 shows an English version of the same page. The document J-1 and the document E-1 have the same description content but only different languages for describing the content.

In the following, a Japanese document having a faithful English translation like Document J-1 or an English document having a faithful Japanese translation like Document E-1 is referred to as a "translated document (group)". Collectively. On the other hand, documents J-2 and J-3 which are only in Japanese, or document E which is only in English
In the following, -2, E-3, E-4, etc. are collectively referred to as "non-translational document (group)".

If all of the documents to be searched are documents with parallel translation, it is possible to easily realize the search of documents in other languages under arbitrary search conditions. That is, for example, in order to be able to search for an English document from Japanese search conditions, the English document corresponding to the Japanese document is set as the search result instead of the Japanese document that matches the Japanese search condition. It should be returned.

However, as shown in the figure, the actual document to be searched includes a document with parallel translation and a document without parallel translation. For at least the latter, the language on the search condition side is set to the language on the search target document side as described above. The conventional method has been to match the language or, conversely, match the language of the document to be searched with the language of the search condition side before performing the search.

On the other hand, in the present invention, as described in detail below, the text of the English document corresponding to the Japanese document that matches the Japanese search condition is regarded as the English search condition, and the English that matches the search condition is used. By returning the document as a search result, without complicated and difficult language conversion processing and statistical processing,
Search English documents from Japanese search conditions.

That is, as schematically shown in FIG. 1, first, a natural sentence input by the operator in Japanese is used as a search condition, and the condition is met (similar to the natural sentence as a whole. Search Japanese documents (in the figure).

Assuming that the Japanese documents J-1 and J-2 meet the above condition, the Japanese document J-1 among these conforming documents has the corresponding English document E-1. Therefore, the English natural sentence which is the text of the English document E-1 is used as a new search condition, and the condition is met (it may be said that it is similar to the natural sentence as a whole).
Search English documents (in the figure).

Then, a document that meets the above new conditions,
That is, the document E-1 itself and the documents E-2 and E-3 similar to the document E-1 are presented to the operator as the final search results.

That is, in the present invention, the original Japanese search condition is not translated into English by itself, but a Japanese document similar to the relevant Japanese document (document J-1 in this example) has already been created manually. Accurate English translation (Document E-
Considering 1) as an English translation of the above conditions for convenience,
Using this as a new search condition, English documents are searched.

The faithful English translation of a Japanese document that matches the Japanese search condition is not the same in content or linguistic as the natural sentence used as the search condition, but similar in content. There is. At least, the semantic / conceptual deviation from the original search condition described in Japanese is smaller than the English translation automatically generated by the conventional machine translation technology, and by performing the search using this as the search condition, The noise contained in the final search result can be reduced.

Further, the manual translation is more natural than the automatically generated translation (that is, it has a quality that the native speaker will not feel uncomfortable to read), and the usage and logic development method is also adapted to the language. It is considered that the similarity to the search target document described in the same language can be more accurately determined by using this as the search condition. However, the translation that serves as a stepping stone does not necessarily have to be created manually, and may be machine-translated if it is not significantly different from the original text. There is also a merit like that).

As described above, according to the present invention, a search for a Japanese document with a natural sentence of Japanese as a search condition input by the operator and a text of an English document corresponding to the Japanese document obtained by the search are performed. It is possible to search for documents written in English from search conditions written in Japanese as a result by performing a chain of search for English documents using natural English sentences as search conditions. Become.

(Embodiment of the Invention) FIG. 2 is an explanatory diagram showing a hardware configuration of a document retrieval apparatus according to an embodiment of the present invention. In the figure, 201 indicates a CPU for controlling the entire apparatus, 202 indicates a ROM storing a basic input / output program, and 203 indicates a RAM used as a work area of the CPU 201.

Reference numeral 204 denotes an HDD (hard disk drive) that controls reading / writing of data from / to an HD (hard disk) 205 under the control of the CPU 201, and 205 denotes an HD that stores written data under the control of the HDD 204. Shown respectively.

Reference numeral 206 denotes an FD (floppy (registered trademark) disk) 207 under the control of the CPU 201.
207 is an FDD (floppy disk drive) that controls the reading / writing of data for
Removable FDs that store written data under the control of FIG.

Reference numeral 208 denotes a cursor, menu, window, or display for displaying various data such as characters and images, and 209 is connected to a network such as a LAN via a communication cable 210, and an interface between the network and the CPU 201. The network I / Fs functioning as are respectively shown.

Reference numeral 211 denotes a keyboard provided with a plurality of keys for inputting characters, numerical values and various instructions.
Reference numeral 2 denotes a mouse for selecting and executing various instructions, selecting a processing target, and moving a cursor.
A CD-ROM 213 is a removable recording medium.
Reference numeral 214 denotes a CD-ROM drive for controlling the reading of data from the CD-ROM 213, and reference numeral 200 denotes a bus or cable for connecting the above-mentioned units.

Next, FIG. 3 is an explanatory diagram showing the functional configuration of the document retrieval apparatus according to the embodiment of the present invention. As shown in the figure, the document retrieval apparatus according to the present invention includes a document storage unit 30
0, Japanese search condition input unit 301, Japanese document search unit 3
02, English search condition generation unit 303, English document search unit 30
4 and a search result display unit 305.

First, the document storage unit 300 is a functional unit that holds a document group to be searched by the Japanese document search unit 302 and the English document search unit 304, which will be described later. Here, it is assumed that the documents in the document storage unit 300 are a large number of WEB documents collected from the Internet, and those written in Japanese are stored in the Japanese document storage unit 300a.
Items written in English are stored in the English document storage unit 300b.

As described above, some of the documents in the document storage unit 300 are documents with parallel translation and the rest are documents without translation. The document with parallel translation has the ID (ID of the parallel translation document described in another language as its attribute information (adjunct information).
However, it may be any information as long as the information can identify the document.

For example, Japanese document J-1 and English document E
If −1 is in a bilingual relationship, the former attribute information has the latter ID “E-1” and the latter attribute information has the former I.
"J-1", which is D, is written in advance.

Next, FIG. 4 is a flowchart showing the procedure of the document search process of the document search device according to the embodiment of the present invention. The functions of the remaining units shown in FIG. 3 will be sequentially described below according to the procedure shown in FIG.

In step S401, the Japanese search condition input unit 301 of the document search apparatus according to the present invention displays the search condition input screen as shown in FIG. 5 on the display 208 and waits for the input from the operator. And the keyboard 21
Search condition input area 500 for characters input from 1 etc.
It will be displayed in sequence inside.

As shown in the figure, it is assumed that a natural sentence consisting of a plurality of sentences is input as a search condition here, but a Japanese sentence such as a natural sentence consisting of a single sentence or a single or a plurality of keywords. Any row may be used.

Next, in step S402, when it is detected that the instruction to execute the search is input, that is, the search execution button 501 shown in FIG. 5 is clicked with the mouse 212 (step S402: Yes), The Japanese search condition input section 301 is the above-mentioned input area 5 at that time.
The character string in 00 is passed as a search condition to the Japanese document search unit 302 described later.

Then, the Japanese document retrieval unit 3 which receives this information
02 searches the Japanese document storage unit 300a according to the above search conditions in step S403. The search method by the Japanese document search unit 302 is such that a document described in a language can be searched from a search condition described in a language (even if it is said that a document search in a single language is possible). Any method may be used, but here, a method generally called "vector space method" is adopted.

In the "vector space method", the cosine distance between the feature vector of the search condition and the feature vector of each document to be searched is calculated, and a document whose distance is absolute or relatively small is calculated. It is presented to the operator as a matching document that matches the search condition.

The feature vector here is an n-dimensional vector consisting of n element values corresponding to n keywords (vocabularies), and each element value is, in the simplest case, the corresponding keyword value. It is determined by the frequency of appearance. For example, the feature vector of a document that includes only one keyword in the body is only the value of the element corresponding to the keyword such as (0, 1, 0, 0, ...) And the remaining n−1.
It is a vector in which all the element values are 0.

According to the vector space method, a document in which the tendency of the keywords appearing in the text as a whole is similar to the search condition has a smaller distance from the search condition, and thus is more likely to be a relevant document. . Therefore, compared to the Boolean search (general keyword search) in which documents are simply selected depending on whether or not a specific keyword in the search condition is included, there is an advantage that there is less noise in the search result.

The Japanese document search unit 302 ranks or scores the documents to be searched based on the above distance, and selects a certain number of documents from the highest rank / highest score documents or a predetermined rank / predetermined score. Documents having a high degree of agreement with the search condition, such as all the above documents, are regarded as conforming documents. Then, the IDs of these matching documents (not limited to IDs, any information that can identify the document) is passed to the English search condition generating unit 303 and the search result display unit 305, which will be described later.

Returning to the procedure of FIG. 4, next step S40.
At 4, the English search condition generation unit 303, which has been passed the search result from the Japanese document search unit 302,
The attribute information of each document specified by D is stored in the Japanese document storage unit 3
Read from 00a. And the ID of the English document in it
Is included, that is, it is determined whether or not at least one of the Japanese documents found by the above search contains a document with parallel translation.

Then, if at least one document with bilingual translation is included in the result (step S404: Yes), then in step S405, the text of the English document corresponding to each bilingual document, that is, The text of the English document specified by each ID in the read attribute information is sequentially read from the English document storage unit 300b. Then, from these English natural sentences, an English document search unit 304, which will be described later,
Generate search conditions to give to.

Note that, here, the English search condition generator 303
In the above, the parallel translations are read out for all the parallel translation documents in the search result by the Japanese document search unit 302, but a part of the parallel translation documents may be selected and only the parallel translations may be read out.

For example, only the documents having the highest matching score with respect to the search condition, such as the documents having the highest rank / highest score among the matching documents to a certain number of documents, or all the documents having a predetermined rank / higher score than the matching documents, The parallel translation is adopted as an English search condition. Conversely speaking, even if there is a bilingual translation, the existence of a bilingual translation is ignored for a document having a low degree of matching with the search condition even if it is a matching document.

As described above, by adopting the parallel translation of the document having a particularly high level among the matching documents, the Japanese search condition input from the Japanese search condition input unit 301 and the English search condition generation unit 303 are generated. The difference from the English search conditions is reduced, and the final search results are more narrowed down and the relevance ratio is high. However, on the other hand, the recall rate will be low, so if you need a complete search, you should use all the parallel translations that exist for Japanese conforming documents as the English search conditions, as in the example here. Good.

When there are a plurality of English documents adopted as the search condition, the English search condition generating unit 303 combines the texts of the respective documents into a continuous natural sentence, and then the English document search unit 304 described later. Hand over to. Then, in response to this, the English document search unit 304 searches for an English document that is similar to this natural sentence as a whole.

However, the text of each adopted document may be regarded as one search condition, and the text may be individually delivered to the English document search unit 304, and the same search may be repeated for the number of documents. . In this case, the search result display unit 305, which will be described later, distinguishes and displays the English documents searched by each condition, or collects the English documents searched by each condition and collectively displays them. It is possible.

It should be noted that the keywords commonly included in each of the documents adopted above (or at least many of them), or the keywords included only in the adopted documents and not included in the other documents. It is also possible to specify only keywords (keywords that characterize the adopted document group with respect to the entire document group that is the parent) and generate search conditions from these keywords.

Next, the English document search unit 304, which has been passed the above search conditions from the English search condition generation unit 303, searches the English document storage unit 300b in step S406, and the English document ID (ID However, any information that can identify the document)
It is delivered to the search result display unit 305 described later.

Similar to the retrieval of a Japanese document by the Japanese document retrieval unit 302, the retrieval of an English document by the English document retrieval unit 304 is based on the vector space method here. However, the method is not necessarily limited to this method, and the search by both functional units does not have to be essentially the same. For example, the Japanese document search unit 302 searches for a document that is substantially similar to a natural sentence that is a search condition by the vector space method described above, and the English document search unit 304 performs a boolean search to find an English document that is adopted as a search condition. You may make it search only the document which certainly contains the keyword in.

Next, the search result display unit 305, to which the IDs of the Japanese documents as the search results and the IDs of the English documents as the search results are delivered from the Japanese document search unit 302 and the English document search unit 304, respectively, In step S407, the search result display screen as shown in FIG. 6 is displayed on the display 208.

In the figure, the Japanese document display frame 6
In 00, each heading of the Japanese document searched by the Japanese document search unit 302 and each heading of the English document searched by the English document search unit 304 are displayed in the English document display frame 601, and the matching degree with respect to the search condition is displayed. They are displayed in descending order. If you click this heading with mouse 212,
The text of the document having the headline can be displayed. Although the document ID is also displayed in parentheses next to the headline in the figure, this may or may not be displayed.

According to the embodiment described above, a document described in a certain language can be searched under a search condition in a different language, but complicated translation processing and statistical processing are performed in the process. Does not occur at all,
The load on the system is much smaller than in the prior art.

Further, it is not necessary to prepare dictionaries for translating between languages, which conventionally takes a lot of time and labor,
However, all that is required is to collect a group of documents that include partially translated documents. And in recent years, including WEB documents,
This collection is easy because there are not a few electronic documents that have been prepared in advance with parallel translations. In other words, the present invention takes note of the fact that the collected document group often includes documents with bilingual translations, and takes advantage of this situation to inspire whether multilingual document retrieval can be realized by using this situation as a foothold. Is.

In the above-described embodiment, the search for the WEB document on the Internet is taken as an example. However, in addition to this, the search for the FAQ document in the internal network of an international company,
The present invention can be applied to searching for academic papers in various languages in various research and educational institutions.

Further, in the above-described embodiment, the Japanese-English document is searched, but it goes without saying that it is also possible to search the English-Japanese document on the contrary. As schematically shown in FIG. 7, for example, an English document having a bilingual relationship is acquired from a Japanese document retrieved in Japanese, and a German document having a bilingual relationship is retrieved from the English document retrieved in the text of the English document. If a German document is acquired and the German document searched by the text of the German document is used as the search result, it means that the German document is eventually searched from the Japanese search condition. It is possible to search for documents in any language from any language by following the expression in which the bilingual documents are translated in multiple languages.

1 and 7, the bilingual document and the document group retrieved by the bilingual document are drawn as if they always exist in the same database. However, it is not always necessary to extract the bilingual document. It is not necessary to perform a search using the parallel translation document for the document group that has become. That is, for example, a parallel translation of a Japanese document searched by the Japanese search condition may be acquired from the database A, and the text may be used as the English search condition to search the other database B.

The above-mentioned Japanese document retrieval unit 302 corresponds to the "first retrieval means" in the claims, and the processing performed by the retrieval section 302 corresponds to the "first retrieval step" in the claims. Further, the English search condition generation unit 303 also serves as the “second search means” and the “generation means” in the claims, and the processing to be performed includes the “second search step” and the “generation step” in the claims. Be done. Also, the English document search unit 304
Corresponds to the "third search means" in the claims, and the process to be performed corresponds to the "third search step" in the claims. Further, the search result display unit 305 corresponds to "display means" in the claims.

The above-mentioned Japanese retrieval condition input unit 30
The 1-search result display unit 305 realizes the function of each unit by the CPU 201 executing instruction processing according to the instruction of the program read from the HD 205 or the like to the RAM 203. Further, in particular, the Japanese document search unit 302 and the English document search unit 304 are specifically realized by the document search engine “ConceptBase Search” manufactured and sold by the applicant.

In addition to the HD 205, the above program can be stored in various recording media such as the FD 207, the CD-ROM 213, or the MO, and can be distributed by this medium or can be distributed via a network. Is.

[0074]

As described above, the invention according to claim 1 is the document retrieval apparatus for retrieving an electronic document described in a second language from a retrieval condition described in a first language. First search means for searching an electronic document described in the first language for electronic documents that match the search condition described in the first language; and an electronic document searched by the first search means. Second search means for searching an electronic document which is a parallel translation of the electronic document described in the second language, and the second document based on the text of the electronic document searched by the second search condition. Generating means for generating a search condition described in the language, and a third search for searching an electronic document matching the search condition generated by the generating means from the electronic document described in the second language. Since the multi-language document search is realized, no complicated work for eliminating the difference in language between the search condition and the search target document is performed, so that the multi-language document search can be performed from any language. An effect is obtained that a document search device capable of performing a search for a document in an arbitrary language by a simple process is obtained.

The invention according to claim 2 is the same as the invention according to claim 1, wherein the first searching means and / or the third searching means has the text of the electronic document as the text. An electronic document that is semantically similar to a natural sentence that constitutes a search condition is searched for as an electronic document that matches the search condition, so only documents that match the overall meaning of the text, such as a field or topic, match the search condition. As a result, there is an effect that it is possible to obtain a document search device capable of accurately searching a document in any language from any language with a simple process.

Further, the invention according to claim 3 is the invention according to claim 1 or 2, wherein the second searching means is all of the electronic documents searched by the first searching means. , The electronic document which is the parallel translation is searched from the electronic document described in the second language. Therefore, the search condition in the second language is the parallel translation of the document obtained in the search in the first language. Documents that are generated and used without omissions, which makes it possible to search documents in any language from any language with simple processing and with high accuracy (specifically, high recall). There is an effect that the device can be obtained.

The invention according to claim 4 is the same as the invention according to claim 1 or 2, wherein the second searching means is one of the electronic documents searched by the first searching means. For some electronic documents, an electronic document that is a parallel translation is searched from the electronic documents described in the second language. Therefore, the search condition in the second language is the first
Among the documents obtained by the search in the language of, for example, only the parallel translations of the documents that have the highest matching degree with the search condition are selectively used to generate, and thereby, the documents of the arbitrary language from the arbitrary language are generated. An effect is obtained that a document search device capable of performing a search with a simple process and with high accuracy (specifically, a high matching rate) can be obtained.

The invention described in claim 5 is the same as the invention described in any one of claims 1 to 4,
Further, since the display means for displaying the information for specifying the electronic document searched by the first searching means and the information for specifying the electronic document searched by the third searching means are provided, it is possible to use the first language. The results of the search and the results of the search in the second language are displayed together on the screen, which allows the results of the multilingual document search, including the intermediate results, to be displayed.
This has the effect of providing a document search device that can be presented to the operator in an easy-to-understand manner.

The invention according to claim 6 is a document retrieval method for retrieving an electronic document described in a second language from a retrieval condition described in a first language, the method being described in the first language. A first search step of searching an electronic document described in the first language for an electronic document that matches the searched search condition, and an electronic document that is a parallel translation of the electronic document searched in the first search step. A second search step of searching for a document from an electronic document described in the second language, and a description in the second language based on the text of the electronic document searched according to the second search condition. And a third search step of searching the electronic document described in the second language for an electronic document that matches the search condition generated in the generating step. Therefore, Even while realizing language document search,
Search condition-No complicated work is performed to eliminate the language difference between search target documents, which allows documents to be searched in any language from any language with simple processing. This has the effect of providing a search method.

According to the invention described in claim 7, there is an effect that a program capable of causing a computer to execute the method described in claim 6 is obtained.

[Brief description of drawings]

FIG. 1 is an explanatory diagram schematically showing the basic principle of the present invention.

FIG. 2 is an explanatory diagram showing a hardware configuration of a document search device according to an embodiment of the present invention.

FIG. 3 is an explanatory diagram showing a functional configuration of a document search device according to an embodiment of the present invention.

FIG. 4 is a flowchart showing a procedure of a document search process of the document search device according to the embodiment of the present invention.

FIG. 5 is an explanatory diagram showing an example of a search condition input screen in the document search device according to the embodiment of the present invention.

FIG. 6 is an explanatory diagram showing an example of a search result display screen in the document search device according to the embodiment of the present invention.

FIG. 7 is an explanatory view schematically showing the basic principle of another embodiment of the present invention.

[Explanation of symbols]

200 buses or cables 201 CPU 202 ROM 203 RAM 204 HDD 205 HD 206 FDD 207 FD 208 display 209 Network I / F 210 communication cable 211 keyboard 212 mouse 213 CD-ROM 214 CD-ROM drive 300 document storage 300a Japanese document storage 300b English document storage 301 Japanese search condition input section 302 Japanese Document Search Department 303 English search condition generator 304 English Document Retrieval Department 305 Search result display area

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Fumiko Kikuchi             3-46 Okihamahigashi, Tokushima City JA Co., Ltd.             Storage system F term (reference) 5B075 KK07 ND03 NK32 NK35 PP24                       PR06 QS01 UU06

Claims (7)

[Claims] 1. A document retrieval device for retrieving an electronic document described in a second language from a retrieval condition described in a first language, wherein the electronic document matches the retrieval condition described in the first language. In the second language, and a first search means for searching the electronic document described in the first language, and an electronic document that is a parallel translation of the electronic document searched by the first search means in the second language. Second searching means for searching from among the retrieved electronic documents, and generating means for generating retrieval conditions described in the second language based on the text of the electronic document retrieved by the second retrieval conditions A third search means for searching an electronic document that matches the search condition generated by the generating means from the electronic document described in the second language, Location. 2. The first search means and / or the third search means searches for an electronic document whose body text is semantically similar to a natural sentence constituting the search condition among the electronic documents. The document search device according to claim 1, wherein the document search device searches for an electronic document that matches the above. 3. The second search means selects, from among the electronic documents described in the second language, an electronic document that is a parallel translation of all the electronic documents searched by the first search means. The document search device according to claim 1, wherein the document search device searches. 4. The second searching means describes, in the second language, an electronic document that is a parallel translation of a part of the electronic documents searched by the first searching means. The document search device according to claim 1 or 2, wherein the electronic document is searched. 5. A display means is provided for displaying information capable of identifying the electronic document retrieved by the first retrieval means and information capable of identifying the electronic document retrieved by the third retrieval means. Claims 1 to 1 characterized
The document search device according to claim 4. 6. A document retrieval method for retrieving an electronic document described in a second language from a retrieval condition described in a first language, the electronic document matching the retrieval condition described in the first language. In an electronic document described in the first language, and an electronic document that is a parallel translation of the electronic document searched in the first search step is described in the second language. A second search step of searching from among the stored electronic documents, and a generating step of generating search conditions described in the second language based on the text of the electronic document searched by the second search conditions. A third search step for searching an electronic document described in the second language for an electronic document that matches the search condition generated in the generating step, and a document search method. . 7. A program for causing a computer to execute the method according to claim 6.