JP2001022787A - Multilinguistic interactive information retrieval system and recording medium where multilinguistic interactive information retrieval program is recorded - Google Patents

Abstract

PROBLEM TO BE SOLVED: To accurately retrieve a document of a language different from a user's retrieval request. SOLUTION: This information retrieval system generates an information retrieval expression by an information retrieval expression generation part 200 at an initial information request made by the user and presents the retrieval result based upon this retrieval expression to the user through a user interface part 100, and an information retrieval result display and selection part 400 makes the user to specify a matching document complying with the user's information request and a requested language. A multilinguistic retrieval expression generation part 500 extracts featured words from the matching document to generate a word list, which is translated into the specified language by using a word list translation part 700; and a multilinguistic information retrieval expression is generated on the basis of the translated word list and on the basis of the retrieval expression, information retrieval is carried out. Consequently, while a language which the user is familiar with is mainly used, a matching document of a different language can be retrieved from an document archive 310 as an object of retrieval.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a multilingual information retrieval system for retrieving a document group composed of a plurality of languages, and more particularly to (1) selecting a document conforming to an information request from a retrieval result of an initial information request. (2) The present invention relates to a multilingual interactive information search system capable of searching for a suitable document described in a specified language by specifying a language of a document to be actually searched.

[0002]

2. Description of the Related Art Here, in response to a user's information request described using a language X, a language Y different from the language X is used.
Is referred to as a "multilingual information search system".

Conventionally, there are roughly two methods for implementing a multilingual information retrieval system. The first method is
This is a method of translating all documents described in languages other than language X included in the document archive to be searched into language X. Further, in the second method, in response to a user's information request described using language X, the information request is translated into language Y, which is the target language specified by the user, and then translated into language Y
This is a method for performing information retrieval by using the following.

[0004]

The first method includes:
(1) It is necessary to translate a document described in a language other than the language X in advance. (2) As a method of executing the translation, there are a manual method and a machine translation method.
The former has a problem that the cost is high, and the latter has a problem that translation accuracy is insufficient. In the second method, the problem in the first method can be solved.
For automatic translation of up to three words, there is a problem that a user's information request input in a very short expression must be appropriately translated into the language Y.

On the other hand, interactive information that presents the result of the initial information search to the user and causes the user to select a suitable document to reconstruct a search request, thereby gradually obtaining a document suitable for the user. There is also a technique called “relevance feedback” that realizes search, but it has not been applied to a “multilingual information search system”.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and has as its object to provide a multilingual interactive information retrieval system and a multilingual interactive information retrieval system capable of accurately retrieving a document in a language different from a user's retrieval request. An object of the present invention is to provide a recording medium on which a multilingual interactive information search program is recorded.

[0007]

To achieve the above object, according to the present invention, a multi-language document group is searched for a document described in a language different from the language of the user's initial information request. In a language interactive information retrieval system, an information retrieval expression generation unit that generates an information retrieval expression based on a user's initial information request input via a user interface unit, and retrieves a document based on the input information retrieval expression An information search execution unit and a document that presents the information search result by the information search execution unit to the user via the user interface unit, and a document that conforms to the user's information request (hereinafter, referred to as a “conforming document”) and a requested document Information search result display selection section that allows the user to specify one language (hereinafter referred to as “designated language”), and a word list in one language in another language. A word list translating unit for translating to a word list to be extracted, and generating a word list by extracting characteristic words in a conforming document; translating the word list into a specified language by the word list translating unit; A multilingual information search formula generation unit for generating an information search formula in a specified language based on the information search formula generated by the multilingual information search formula generation unit, and executing the information search execution unit. Suggest.

According to a second aspect of the present invention, there is provided the multilingual interactive information retrieval system according to the first aspect, further comprising a document information acquisition section for acquiring word frequency distribution information of a matching document from a document group, and the multilingual information retrieval system. The formula generation unit proposes a feature that generates a characteristic word list in a matching document based on the word frequency distribution information obtained from the document information acquisition unit.

Further, the present invention proposes a recording medium on which a multilingual interactive information search program for realizing the multilingual interactive information search system according to claims 1 to 2 by executing the program is recorded.

According to the present invention, the language X that the user is good at
When an initial search is executed in response to an information request in a native language (such as a native language), an information search is performed from a group of documents using this language X, and a search result is presented via a user interface unit. The user selects a matching document (matching document) from the group of documents of the search result described in the language X, and
When the language Y of the document to be actually searched is specified, a characteristic word list is generated in the matching document, the word list is translated into the language Y, and information retrieval is performed from the document group based on the translated word list. As described above, it is possible to search for a suitable document in a different language Y while mainly using the language X that the user is good at.

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENTS A multilingual interactive information retrieval system according to one embodiment of the present invention will be described with reference to FIG. FIG. 1 is a configuration diagram of a multilingual interactive information search system according to the present embodiment.

As shown in FIG. 1, the multilingual interactive information retrieval system includes a user interface 1
00, the information search formula generation unit 200, and the information search execution unit 3
00, a document archive 310, an information search result display selection unit 400, a multilingual information search expression generation unit 500, a document information acquisition unit 600, and a word list translation unit 700.

The user interface unit 100 controls the interaction with the multilingual interactive information retrieval system of the present invention. That is, the user inputs an initial information request, displays a result of information search execution, specifies a suitable document by the user, and selects a language of a document to be actually searched.

The information search formula generation unit 200 interprets the user's initial information request input via the user interface unit 100, and enables the information search execution unit 300 to actually execute the information search process for the document archive 310. A new information search formula is generated, and the generated information search formula is sent to the information search execution unit 300.

The information retrieval execution section 300 receives information retrieval expressions generated by the information retrieval expression generation section 200 and the multilingual information retrieval expression generation section 500 as input, and records information recorded in a plurality of languages constructed in advance. The document archive 310 to be searched is searched, and the search result is sent to the information search result display selection unit 400. In the present invention, since the information search execution unit 300 and the document archive 310 to be searched for can use an existing search engine or the like, the details are not specified in this specification.

The information search result display selection unit 400 sends information of the information search result by the information search execution unit 300 to the user interface unit 100. Further, the user interface unit 100 stores the ID of the conforming document selected by the user from the information retrieval result and the language name of the document to be actually retrieved.
Receive from The received information is sent to the multilingual information search formula generation unit 500.

The multilingual information search expression generation unit 500 is configured to output the I of the conforming document transmitted from the information search result display selection unit 400.
With the D number and the language name (designated language) of the document to be actually searched as input, a multilingual information search formula for searching for a compatible document in the designated language is generated. At this time, ID
The word frequency distribution information in the document specified by the number is obtained by the document information obtaining unit 600. Usually, such a function is provided in a search engine used as the information search execution unit 300, or can be easily realized as an additional function. In generating a multilingual information search formula, a process of translating a word list in a source language into a word list in a designated language (target language) is performed.
By calling. Note that the word list translator 7
00 can be realized by, for example, the one proposed by the applicant in Japanese Patent Application No. 11-153441, “Translation word selection dictionary automatic generation device and automatic translation device, and a medium recording these programs” and the like. Details are not specified. The generated multilingual information search formula is sent to the information search execution unit 300, and a search for a matching document in the designated language (target language) is performed.

An example of the multilingual interactive information search system according to the present embodiment will be described below with reference to FIGS. FIG. 2 is a flowchart illustrating the operation of the multilingual interactive information search system according to the present embodiment, and FIGS. 3 and 4 are diagrams illustrating an example of a screen image of the user terminal generated by the user interface unit.

First, the user's initial information request input via the user interface unit 100 is "ATM", "
Network ".

The information search formula generation unit 200 is based on the input initial information request of the user and executes the information search execution unit 300
Generates an information search formula that conforms to the query format of the search engine used as a search engine (step S
1). As described above, in this specification, the details of the information search execution unit 300 are not specified, but in the following description, it is assumed that the information search execution unit 300 accepts an unordered list of words as an inquiry format. .
Therefore, the information search formula generation unit 20
The information retrieval formula generated by 0 is ("ATM", "network"). Note that in an information search system such as a normal search engine, it is often allowed to combine words constituting an inquiry with logical operators such as and, or, and not. It should be noted that the present invention does not limit such variations of the information retrieval formula.

Next, the information search execution unit 300 searches the document archive 310 based on the input information search formula (step S2).

Next, the information search result by the information search execution unit 300 is sent to the user interface unit 100 via the information search result display selection unit 400, and the search result is presented to the user, and the search result by the user is displayed. , A conforming document and a designated language (target language) are input (step S3).

FIG. 3 shows the user interface unit 1
9 shows an example of a screen image of the user terminal generated by the user terminal 00. One of the important items illustrated in FIG. 3 is that information such as the title of a document in the document archive 310 searched by the information search execution unit 300 using the information search formula is presented to the user. Is to be done. It should be noted that these pieces of information are essential when the user selects a suitable document, but the present invention does not limit the type of information presented to the user.
The second important thing is to specify, on the screen, a means for selecting a conforming document actually selected by the user (in FIG. 3, realized as a check box) and a language of the document to be actually searched. (In FIG. 3, realized as a menu). These two items are essentially important, and the present invention does not specify the details of the screen design displayed on the user terminal.

FIG. 4 shows a state in which the user has specified a conforming document and the language of the document that the user actually wants to search for the search results illustrated in FIG. As shown in FIG. 4, the user selects three documents of document ID: 101, document ID: 5, and document ID: 280 as conforming documents, and selects "English" as a designated language (target language). Shall be done.

Next, the information selected by the user is transmitted to the multilingual information search generator 50 via the information search result display selector 400.
0, and the multilingual information search generator 500 performs a search in the designated language (step S4). In the example of FIG. 4, when the user presses the “feedback search execution” button, these pieces of information are sent to the multilingual information search generation unit 500 via the information search result display selection unit 400.

The operation of the multilingual information retrieval formula generation unit 500 in step S4 will be described below.

First, the multilingual information search formula generation unit 500 that has received the information issues a request to the document information acquisition unit 600, and acquires word frequency distribution information in the designated conforming document (step S41). .

Next, based on the word frequency distribution information acquired from the document information acquiring section 600, a word group characterizing those documents is extracted (step S42).

Next, from the word group extracted from each document,
The word list translation unit 700 generates a word list to be translated (step S43).

Next, the word list is sent to the word list translating unit 700, and a translation result translated into a specified language (target language) is received (step S44).

Next, the information retrieval execution unit 3 is searched from the word list of the translation result in the designated language (target language).
00 generates an information search formula that can be accepted (step S45).

Finally, the above-described information search formula is sent to the information search execution unit 300, and the received search results are sent to the user interface unit 100, thereby presenting the search results in the designated language (target language) to the user (step). S46).

Hereinafter, the operation of each step of the multilingual information retrieval formula generation unit 500 will be described more specifically with reference to the example of FIG.

First, step S41 will be described with reference to FIG. FIG. 5 is a diagram illustrating an example of word frequency distribution information in a specified document obtained by the document information obtaining unit 600. As shown in FIG. 5, the word frequency distribution information in each document is obtained by adding a real value indicating the importance of the word in the document to each word included in the document. As an index for determining the importance of a certain word in a certain document, an appearance frequency can be simply considered, but a value called a tf * idf method is often used in text information search. In the present invention, it is assumed that the information search execution unit 300 employs such a standard method, and details thereof are not specified.

Next, in step S42, a word group characterizing a certain document is extracted. Tf * idf described above
If the importance of a word is given by a method such as a method, for example, a word group characterizing a document can be extracted by selecting N words in descending order of this value. Here, for simplicity of explanation, it is assumed that up to five words are selected from each document in descending order of value. Here, since the value of N should be experimentally determined in consideration of the characteristics of the document archive to be searched, the method of determining it is not specified in the present invention. In FIG. 5, words selected by such a method are indicated by shading.

Next, in step S43, the word groups extracted from each document are combined to extract a word group representing the characteristics of the document group selected as a suitable document by the user. These words are words to be translated into a designated language (target language) selected as a language of a document actually searched by the user. M selected by user
Since a maximum of N words are extracted from each (for example, three) matching documents, the number of words to be translated is a maximum of M * N.
If the number of words increases, the recall rate in the search can be reduced (reduction of omissions in the search), but on the contrary, the relevance rate may be low (there are many non-conforming ones in the search). There is. Therefore, it is desirable to select an appropriate number of words. Although the number of such words should be determined experimentally in consideration of the characteristics of the document archive 310 and the like, it is not defined in the present invention.
For all candidate words x, the value W
(X) may be calculated, and L values may be selected in descending order of this value.

[0037]

(Equation 1)

Here, weight (x, k) is the importance of word x in document k. If word x does not exist in document k, weight (x, k) = 0.
Df (x) is the number of documents including the word x in the M documents selected by the user. FIG. 6 shows the result of calculating W (x) for each selected word shown in FIG. Here, if L = 6, for example, the shaded word group in FIG. 6 is selected.

Next, in step S44, a word list ("network", "ATM", "hub", "LA") composed of the word group extracted in step S43 is used.
N "," switch "," FDDI ") to the word list translator 700 to receive the translation result in the specified language (target language). The translation result word list is, for example, (" network "," ATM "). , "Hu
b "," LAN "," switch "," FDDI ")
become that way.

Next, in step S45, a multilingual information retrieval formula is generated from the word list in the designated language (target language) obtained in step S44. Here, as the multilingual information search formula, as in the case of the information search formula in the source language, ("network", "AT
M "," hub "," LAN "," switch ","
FDDI ") as an unordered list of simple words.

Next, in step S46, the multilingual information search expression generated in step S45 is sent to the information search execution unit 300, and a search result for a target language document specified by the user is received. The received search result is sent to the user interface unit 100, and the search result is presented to the user.

As described above, in the multilingual interactive information search system according to the present embodiment, when an initial search is executed in response to an information request in a language X (such as a native language) that the user is good at, the language X is used. Information retrieval is performed from the document group, and the retrieval result is presented via the user interface unit 100. When the user selects a matching document (matching document) from the group of documents of the search result described in the language X and specifies the language Y of the document to be actually searched, a characteristic word list is generated in the matching document. And
This word list is translated into the language Y, and information retrieval is performed from the document archive 310 based on the translated word list. As described above, it is possible to search for a suitable document in a different language Y while mainly using the language X that the user is good at.

In this embodiment, the document archive 310 containing a large number of documents is exemplified as a document group having a plurality of languages to be searched, but the present invention is not limited to this. That is, the document to be searched may be recorded at a local location or may be distributed and recorded at a global location. For example, all documents accessible on the Internet may be targeted. In this case, the search execution unit and the document information acquisition unit need to be able to access each document.

Further, in the present embodiment, the case where the user has only once performed the selection input of the conforming document and the designated language after the initial information request has been described, but the present invention is not limited to this. is not. After a search result in a specified language (target language) is presented using a multilingual information search formula, a search target may be narrowed down by further selecting a suitable document and a specified language based on the search result. That is, in the flowchart in FIG.
Relevant documents may be narrowed down by repeating steps S3 and S4 a plurality of times.

Further, in the present embodiment, the means for generating a word group characterized in a specified document acquires word frequency distribution information in the document and generates a word group based on the word frequency distribution information. However, the present invention is not limited to this. For example, when each document in a document group has a keyword that characterizes the document as a keyword, the keyword may be acquired.

[0046]

As described in detail above, according to the multilingual interactive information retrieval system of the present invention, a document to be searched for a suitable document in a different language Y while mainly using the language X that the user is good at. It becomes possible to search from a group.

More specifically, for example, in the example used in the description, the search result for the initial information request in Japanese (“ATM”, “network”) is, as shown in FIG. It is expected that documents on ATMs related to ATMs and documents on ATMs as automatic teller machines will coexist. However, if the user can appropriately select a suitable document as shown in FIG. 3, the search expression for the feedback search in English, which is the target language, is (“network”) as described above.
k "," ATM "," hub "," LAN "," swi
tch "," FDDI "), and AT in the former sense
It can be expected to search English documents including M.

When generating a search expression in English, “hub” means “hub (meaning of relay)”.
"Habu (meaning snakes living in Okinawa)" is translated, but "LAN", "network", "FDD"
By forming and translating a list with words such as "I", it is expected that "hub" will be correctly selected as a translated word, that is, search accuracy in multiple languages can be improved.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a multilingual interactive information search system according to an embodiment;

FIG. 2 is a flowchart for explaining the operation of the multilingual interactive information search system according to the embodiment;

FIG. 3 illustrates an example of a screen image of a user terminal generated by a user interface unit.

FIG. 4 is a diagram illustrating an example of a screen image of a user terminal generated by a user interface unit.

FIG. 5 is a diagram showing an example of word frequency distribution information in a specified document obtained by a document information acquisition unit.

FIG. 6 is a diagram showing an example of a word group characterized in a designated document;

[Explanation of symbols]

100: User interface unit, 200: Information search formula generation unit, 300: Information search execution unit, 310: Document archive, 400: Information search result display selection unit, 500: Multilingual information search formula generation unit, 600: Document information acquisition Part, 700
… Word list translator

Claims (4)

[Claims] 1. A multilingual interactive information retrieval system for retrieving a document described in a language different from a language of a user's initial information request from a group of documents composed of a plurality of languages, wherein a user input via a user interface unit is provided. An information search formula generation unit that generates an information search formula based on an initial information request of the user, an information search execution unit that searches for a document based on an input information search formula, and a user interface that provides information search results by the information search execution unit Information that is presented to the user via the section and allows the user to specify a document that conforms to the user's information request (hereinafter referred to as “conforming document”) and the language of the requested document (hereinafter referred to as “designated language”) A search result display selection unit; a word list translation unit for translating a word list in one language into a word list in another language; A multilingual information search for extracting a characteristic word in a word list, generating a word list, translating the word list into a specified language by the word list translating unit, and generating an information search formula in the specified language based on the translated word list A multilingual interactive information search system, comprising: a formula generation unit, wherein the information search execution unit is executed by an information search formula generated by the multilingual information search formula generation unit. 2. A document information acquisition unit for acquiring word frequency distribution information on a conforming document from a document group,
2. The multilingual information search formula generating unit generates a characteristic word list in a matching document based on word frequency distribution information obtained from the document information obtaining unit.
The described multilingual interactive information retrieval system. 3. A recording medium storing a multilingual interactive information retrieval program for retrieving a document described in a language different from the language of an initial information request of a user from a group of documents composed of a plurality of languages, wherein the program comprises: Generating an information search formula based on the user's initial information request input via the interface unit; searching for a document based on the generated information search formula; and transmitting the search result via the user interface unit Presenting to the user, causing the user to specify a conforming document and a designated language from among them, extracting characteristic words in the conforming document to generate a word list, and translating the generated word list into the designated language Generating an information search expression in a specified language based on the translated word list; Retrieving a document based on the generated information retrieval formula in the designated language. 4. The step of generating the word list includes the steps of acquiring word frequency distribution information of a matching document from a group of documents; and generating a characteristic word list of the matching document based on the acquired word frequency distribution information. A recording medium recording a multilingual interactive information search program, characterized by having a step.