JP2003242170A - Document search device, document search method, and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a document search device, a document search method, and a recording medium capable of suppressing quality deterioration of document ranking regardless of the quality of search request. <P>SOLUTION: From a document group stored in a document storage means such as a hard disk 4, a document where an initial search language group stored in a search language storage means such as a hard disk 4 appears, is ranking-searched to acquire a first ranking. An enlarged language group is generated from a document group included therein. A document where the initial search group and the enlarged language group appear is ranking-searched to acquire second ranking. A control part such as CPU 22 and a memory 3 which adjusts the first ranking and the second ranking to acquire a new third ranking is provided. <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 search device, a document search method and a recording medium for searching and ordering documents in which a specified search word group appears from a document group.

[0002]

2. Description of the Related Art Conventionally, when a document is searched by a plurality of words (search word: keyword) designated as a search condition, the degree to which each document in a document group meets the search condition ( Hereinafter, various document search apparatuses have been proposed (Japanese Patent Laid-Open No. 2000-322416, Japanese Laid-Open Patent Publication No. 10-199810) that obtains the goodness of fit and orders the documents in the descending order of the goodness of fit (hereinafter also referred to as ranking). -4
9549).

In this type of document retrieval apparatus, a retrieval request is represented by a retrieval word group {Ti}, and the matching degree between the document D and the retrieval request is the sum of scores obtained for each retrieval word Ti for the document D. It is often defined by. Here, the score is a numerical value indicating the degree of importance of the document D, and is a numerical value by setting a predetermined criterion between the document D and the search condition. It should be noted that this digitization also includes weighting by the sentence or word to be searched.

When outputting the result of such a document search, the above-mentioned search request is ranked according to the matching degree of each document in the searched document group, and the quality of the document ranking is average. It is evaluated by the precision rate (AVP). Here, the average relevance ratio is the ratio of relevant documents (documents that satisfy the search request) included in the top r document groups in the ranking for r = 1, 2, ...
It is the average of the individual values.

Further, as one method for obtaining a high quality document ranking, a pseudo conformance feedback method is used.
(pseudo-relevance feedback method) is available. In this pseudo relevance feedback method, a search (initial search) is performed using a search word group (initial search word group) expressing a search request, and a document group (seed document group) ranked higher in the search result is searched. A new search word (extended word group) included is generated. The result of the search (extended search) using both the initial search word group and the expanded word group is the final output.

On the other hand, a function called related word search has been conventionally proposed. Here, the related word search is to obtain another search word (related word) group related to the given search word group. For example, a related word group such as "dielectric constant" or "polarization" is obtained for the search word "hysteresis". Such related word search is often used for document search support, such as presenting a new search word group to the searcher for the search word group given by the searcher. The related word search can be realized by a method similar to the pseudo compatibility feedback method. That is, the expanded word group may be output as the related word group.

[0007]

However, in such a conventional document search apparatus, when the pseudo compatibility feedback method is used, an expanded word group is generated by the initial search word group,
Since the expanded search is performed using both of them, if the given search request (initial search word group) is not appropriate, the quality of the initial search results is low, and the expanded word obtained from the seed document is the relevant search request. There are many cases that are unrelated to, and there is a problem that the quality of the extended search results deteriorates. Further, when using the pseudo compatibility feedback method, since the quality of the search request is determined by the searcher, and it is very difficult for the document search device to evaluate the quality of the quality,
There is a problem in that the expanded search may result in a search result that is inferior in quality to the initial search result. Further, when the related word search is realized by a method similar to the pseudo relevance feedback method, the obtained expanded word group is, so to speak, treated as a candidate of the final related word group. There is a problem that it is not easy to select only appropriate ones from the candidates.

The present invention has been made to solve such a problem, and is capable of suppressing the quality deterioration of the document ranking regardless of the quality of the search request.
A document search method and a recording medium are provided.

[0009]

A document retrieval apparatus according to claim 1 of the present invention stores a document storage unit that stores a document group to be ranked for searching, and a first search word group used for ranking search. The search word storage means and the document group stored in the document storage means are searched for a document in which the first search word group stored in the search word storage means appears, and the first ranking is performed. A document in which the second search word group is acquired from the document group included in the first ranking, and the first search word group and the second search word group appear from the document group stored in advance. A ranking acquisition unit that acquires a second ranking by performing a ranking search, adjusts the second ranking according to the first ranking, and acquires a third ranking.

With this configuration, the ranking (first ranking) of the documents in which the initial search word group appears is obtained, the extended word group is generated from the document group included in the first ranking, and the initial search word group and the extended word group are expanded. Even if the quality of the search request by the initial search word group is low, the ranking (second ranking) of the document in which the word group appears is obtained and the second ranking is adjusted based on the first and second rankings. Therefore, it is possible to realize the pseudo compatibility feedback method with less quality deterioration than the conventional one. Generally, in the pseudo-adaptation feedback method, parameters used in the ranking search are adjusted so that the quality of the upper part of the ranking is particularly good in the initial search and the quality of the entire ranking is good in the extended search. . Therefore, by adjusting the second ranking by the above-mentioned first ranking, the two are merged, and the result in which the quality of the upper portion is good and the quality of the whole is also good is obtained.

According to a second aspect of the present invention, there is provided a document search method in which a first preset document group is selected from a previously stored document group.
A first step of obtaining a first ranking by searching the documents in which the search word group appears, and a second step of generating a second search word group from the document group included in the first ranking. And performing a ranking search for documents in which the first search word group and the second search word group appear from the previously stored document group, and obtaining a second ranking.
And a step of adjusting the second ranking according to the first ranking to obtain a third ranking.
And the steps of.

According to this method, a new ranking (third rank) is obtained from the first ranking acquired using the initial search word group and the second ranking acquired using the initial search word group and the expanded word group. As described above, even if the quality of the search request by the initial search word group is low as described above, it is possible to realize the pseudo compatibility feedback method in which the quality deterioration is less than in the conventional case.

Further, the recording medium according to claim 3 of the present invention performs a ranking search for a document in which a preset first search word group appears in a computer from a document group stored in advance, and the first search word group is searched. A first step of obtaining a ranking; a second step of generating a second search word group from a document group included in the first ranking; and a first search word from the previously stored document group Group and a document in which the second search word group appears, are searched for a ranking, and a third step of obtaining the second ranking and adjusting the second ranking by the first ranking to adjust the third ranking The fourth step of acquiring and a program for executing the fourth step are recorded.

With this configuration, a new ranking (third rank) is obtained from the first ranking acquired using the initial search word group and the second ranking acquired using the initial search word group and the expanded word group. It is possible to easily obtain the program for obtaining the ranking) and realize the pseudo compatibility feedback method in which the quality deterioration is less than in the conventional case even when the quality of the search request by the initial search word group is low as described above.

A document retrieval apparatus according to claim 4 of the present invention is
From a document storage unit that stores a document group to be searched for a ranking, a search word storage unit that stores a first search word group used for a ranking search, and a document group stored in the document storage unit, A document in which the first search word group stored in the search word storage unit appears is searched to obtain the first ranking, and the document group included in the first ranking is changed to the first search word group. A related related word group is generated, a document in which a related word included in the related word group appears is searched from the document group stored in the document storage unit to obtain a second ranking, and the first ranking is acquired. And a related word evaluation means for evaluating each related word included in the related word group according to the ranking and the second ranking.

With this configuration, in the related word search by a method similar to the pseudo compatibility feedback method, the first ranking obtained by using the initial search word group and the second ranking obtained by using the related word group. From this, since the evaluation of each related word is performed, for example, by calculating the average precision rate AVP using the initial search result, the quality of the related word can be appropriately evaluated.

According to a fifth aspect of the present invention, there is provided a document search method in which a first preset document group is selected from a previously stored document group.
Search for a document in which the search word group appears, and obtain a first ranking; and a related word group related to the first search word group from the document group included in the first ranking. And a third step of performing a ranking search for a document in which a related word included in the related word group appears, from the prestored document group, and obtaining a second ranking. And a fourth step of evaluating each related word included in the related word group according to the first ranking and the second ranking.

According to this method, the step of evaluating each related word from the first ranking acquired using the initial search word group and the second ranking acquired using the related word group as described above is performed. Therefore, it is possible to appropriately evaluate the quality of the related word in the related word search by the method similar to the pseudo suitability feedback method.

Further, the recording medium according to claim 6 of the present invention performs a ranking search for a document in which a preset first search word group appears in a computer from a document group stored in advance, and a first search is performed. A first step of obtaining a ranking; a second step of generating a related word group related to the first search word group from a document group included in the first ranking; and the previously stored document group From, a ranking search is performed for documents in which related words included in the related word group appear,
A program for executing a third step of obtaining a second ranking, and a fourth step of evaluating each related word included in the related word group by the first ranking and the second ranking. It has the recorded configuration.

With this configuration, a program for evaluating each related word from the first ranking acquired using the initial search word group and the second ranking acquired using the related word group as described above is provided. The quality of the related words can be appropriately evaluated in the related word search by a method similar to the pseudo compatibility feedback method, which is easily acquired.

[0021]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. [First Embodiment] FIG. 1 shows a hardware configuration of a document search apparatus according to a first embodiment of the present invention. In FIG. 1, the document search apparatus 100 includes a C that controls the entire apparatus.
A PU 22, a memory 3 including a ROM and a RAM for storing programs and necessary data for implementing various functions under the control of the CPU 22, a search target document, search conditions, search results, and the like are stored. Hard disk 4 for inputting, an input unit 5 for inputting necessary instructions and data with a pointing device such as a keyboard and a mouse, an output unit 6 including a CRT or a liquid crystal display, and writing of data to a flexible disk ( A flexible disk drive 7 for reading and updating and a CD-ROM drive 8 for reading data from a compact disk read-only memory (CD-ROM) are provided, and each unit 3 to 8 and 22 is connected by a bus 9. Has been done.

FIG. 2 shows a functional configuration of the document search device according to the first embodiment of the present invention. In FIG. 2, the document storage unit 14 stores an attribute group including a document to be searched and a document number given to this document, and this function is realized by the hard disk 4 or the like. The search word buffer 11 stores the search word group input by the input unit 5, and this function is realized by the hard disk 4 or the like. The index unit 15 stores, for each of the index words, a group of attributes including a document number of a document stored in the document storage unit 14 and in which the index word input by the input unit 5 appears. This function is realized by the hard disk 4 or the like. The search unit 13 uses the document storage unit 14 for the search word group input by the input unit 5.
Among the documents stored in, the document number and the score of this document are obtained for the document in which the search word group appears, and a list including the document number and the score as an element is generated. The function is realized by the CPU 22, the memory 3, and the like. The ranking buffer 1 and the ranking buffer 2 store a list of pairs of document numbers and scores generated by the search unit 13, and this function is realized by the hard disk 4 or the like.

The control unit 12 uses the search unit 13 to perform a ranking search for the document group stored in the document storage unit 14 using the search word group stored in the search word buffer 11 and to perform the ranking search. A list of document number and score pairs of documents in which a group appears is stored, this list is stored in the ranking buffer 1, and the document numbers constituting the elements of the list stored in the ranking buffer 1 correspond to this document number. Document to be acquired from the document storage unit 14, a new search word group is generated from the acquired document group, and the new search word group and the search word group stored in the search word buffer 11 are combined to create a new search word group. A search word group is formed, and a search list 13 is used to obtain a list of pairs of document numbers and scores by using this new search word group.
The list stored in the ranking buffer 1 is adjusted by the list stored in the ranking buffer 1, and the adjusted list is output by the output unit 6. This function is performed by the CPU 22, the memory 3, and the like. Will be realized.

Next, the document search method according to this embodiment will be described with reference to FIG. The control unit 12 acquires the search word group (initial search word group) stored in the search word buffer 11 (step S101), and causes the search unit 13 to perform a document ranking search using the initial search word group. As a result of the initial search, a list having a set of document number and score as an element is acquired, and the obtained ranking (first ranking:
The initial search result) R1 is stored in the ranking buffer 1 (step S102).

Next, the control unit 12 sorts the rankings stored in the ranking buffer 1 in the order of scores to obtain the upper n document number groups, and the document (seed document) corresponding to each document number of this document number group. ) Are acquired one by one from the document storage unit 14, and a search word group (extended word group) is generated from each seed document (step S103). That is, the seed document is divided into words, and the document frequency of the words is obtained. Here, the document frequency of the word w is the number of seed documents in which the word w appears. Then, the word groups are ranked according to the document frequency,
Search words (extended words) by selecting words from the top as many as you like
And

Next, the search unit 13 executes a document ranking search using the initial search word and the expanded word, and acquires a list having a set of a document number and a score as an expanded search result, which is obtained. The ranking (second ranking: expanded search result) R2 is stored in the ranking buffer 2 (step S104).

Next, the control unit 12 acquires a new ranking (third ranking) R3 from the ranking R1 stored in the ranking buffer 1 and the ranking R2 stored in the ranking buffer 2 (step S105). . This work is executed using the ranking buffer 2. Specifically, the ranking R1 and the ranking R2 are sorted in the order of scores. Then
The element e is taken out in order from the ranking R2, and the following processing is performed. First, when the extracted element e is the upper r1 element of the ranking R1 and the upper r2 element of the ranking R2, the score S (e) is obtained by the following equation (1). S (e) = α × r1 + (1-α)
Xr2 (1) However, α is a constant of 0 or more and 1 or less, which is a hybrid parameter between rankings.

Next, the score of the element e in the ranking R2 is replaced with S (e), and the ranking R2 is sorted in the ascending order of the new scores. Thus, the new ranking R3 is acquired. With respect to the new ranking R3 thus acquired, the output unit 6 outputs the document number or the document number and the score stored in the ranking buffer 2 (step S106).

When the document ranking search is performed by the search unit 13 as described above, the search unit 13 uses the index unit 1
5, the document number and the score S of the appearing document D in which the search word T (the above-mentioned initial search word, or the initial search word and the expanded word) given from the control unit 12 appear are used.
Find (T, D). The score S (T, D) is the appearance frequency Tf of the search word T in the document D (for the document D,
Tf / Df based on the number of times the search word T appears and the document frequency Df for the collection of the search word T (the number of times the document D including the search word T appears for the search word T).
Is a numerical value given as.

As described above, in the document search device 100 according to the first embodiment of the present invention, the initial search word group stored in the search word buffer 11 appears from a plurality of documents stored in the document storage unit 14. To obtain the first ranking, generate an extended word group from the document group included in the first ranking, and perform a ranking search for documents in which the initial search word group and the extended word group appear. Control unit 1 that obtains the second ranking, adjusts the second ranking according to the first ranking, and obtains a new third ranking
Since 2 is provided, even if the quality of the search request by the initial search word group is low, it is possible to realize the pseudo compatibility feedback method with less quality deterioration than in the past.

Further, the document search method according to the first embodiment of the present invention performs a ranking search for a document in which a preset initial search word group appears from a plurality of documents stored in advance,
Step S102 of acquiring the first ranking, and
Step S103 of generating an expanded word group from the document group included in the ranking, and a step of performing a ranking search for a document in which the initial search word group and the expanded word group appear from the plurality of stored documents to obtain the second ranking. S104
And the step S105 of obtaining the third ranking by adjusting the second ranking according to the first ranking, even when the quality of the search request by the initial search word group is low, compared to the conventional case. A pseudo-fitness feedback method with less quality deterioration can be realized.

In the first embodiment, the case in which the program for implementing the above-mentioned document retrieval method is stored in the memory 3 of the control unit 12 has been described. A computer-readable recording medium (CD-ROM, FD, magneto-optical disk (MO), in which a program for realizing the search method is recorded,
Even if a mini disk (MD) or a rewritable CD (including CD-RW) is attached to the document retrieval apparatus 100 and this recording medium is read by the CD-ROM drive 8 or the like to execute the program, the same effect is obtained. Is obtained. With this configuration, the program can be easily updated by moving and exchanging the recording medium.

Further, in the first embodiment, the case in which the program for realizing the above-mentioned document retrieval method is stored in the memory 3 of the control unit 12 has been described, but the present invention is not limited to this. The same effect can be obtained by providing the document retrieving apparatus 100 with the above and downloading the program from the external apparatus on the network such as LAN to the control unit 12 and executing the program. With this method, the program can be easily updated via the network.

[Second Embodiment] The hardware configuration of the document retrieval apparatus according to the second embodiment of the present invention is almost the same as that of the first embodiment. Are assigned the same reference numerals and description thereof will be omitted. Also,
The functional configuration according to the present embodiment is substantially the same as that of the first embodiment except for the related word buffer 16, and thus the same reference numerals are given to the same components and the description thereof will be omitted.

FIG. 4 shows the functional arrangement of a document search device according to the second embodiment of the present invention. This is different from the first embodiment in that a related word buffer 16 is further provided, a document in which an initial search word group appears is searched for a ranking to obtain a first ranking, and a document group included in this first ranking is acquired. A related word group is generated, the documents in which the related words included in this related word group appear are searched for to obtain the second ranking 2 for each related word, and the related words are identified by the first and second rankings. The difference is that the related words are evaluated and stored in the related word buffer 16 according to the evaluation value.

In FIG. 4, the control unit 12 causes the search unit 13 to perform a ranking search for the search word group stored in the search word buffer 11, and sets the combination of the document number and the score of the document in which the search word appears. A list is obtained and stored in the ranking buffer 1, and according to the document numbers constituting the elements of the list stored in the ranking buffer 1,
A document corresponding to the document number is acquired from the document storage unit 14, a related word group related to the search word group is generated from the acquired document group, and a related word configuring the related word group is generated by the search unit 13. A ranking search is executed for each of these, and a list of document number and score pairs of documents in which this related word appears is stored in the ranking buffer 2 and the list stored in the ranking buffer 1 and the ranking buffer 2 The related word is evaluated, the related word is stored in the related word buffer 16 according to the evaluation result, and the output unit 6 outputs the related word group. This function is performed by the CPU 22, the memory 3, and the like. Will be realized. The related word buffer 16 stores a group of related words after evaluation, and this function is realized by the hard disk 4 or the like.

Next, the document search method according to this embodiment will be described with reference to FIG. The control unit 12 acquires the search word group (initial search word) stored in the search word buffer 11 (step S201), causes the search unit 13 to perform a document ranking search using the initial search word, and performs the initial search. A list having, as elements, a document number and a score of a document in which a word appears is acquired, and the acquired ranking (first ranking) is stored in the ranking buffer 1 (step S202).

Next, the rankings stored in the ranking buffer 1 are sorted in the order of scores to obtain a group of the top n document numbers, and the documents (seed documents) corresponding to the respective document numbers are obtained one by one from the document storage unit 14. A search word group (here, a related word candidate group is shown) is generated from each seed document (step S203). That is, the seed document is divided into words, and the document frequency of the words is obtained. Here, the document frequency of the word w is the number of seed documents in which the word w appears. Furthermore, the word groups are ranked according to the document frequency described above,
A desired number of words w are selected from the upper order to be related word candidates.

Next, the related words are extracted one by one from the related word candidate group and stored in the related word buffer 16 when the following conditions are satisfied (steps S204 to S207).
That is, first, the rankings in the ranking buffer 1 are sorted in the order of scores, and the top x items are regarded as a set R of relevant documents for the initial search word. Next, a document ranking (second ranking) C is obtained by using the related word candidate c extracted from the above-mentioned related word candidate group as a search word, and the average matching rate of the document ranking C is further obtained using the set R of the matching documents. (A
VP). When the average precision ratio exceeds a predetermined threshold value, the related word candidate c is set to the related word buffer 16
To store. When the evaluation of all the related word candidates c is completed (step S208), the output unit 6 outputs the related word group stored in the related word buffer 16 (step S209).

As described above, in the document search apparatus 100 according to the second embodiment of the present invention, the initial search word group stored in the search word buffer 11 appears from the documents stored in the document storage unit 14. To obtain a first ranking, and generate a related word candidate group related to the initial search word group from the document group included in the first ranking. From the plurality of stored documents A control in which a document in which a candidate included in the related word candidate group appears is ranked to obtain a second ranking, and each candidate included in the related word candidate group is evaluated by the first and second rankings. Since the unit 12 is provided, it is possible to appropriately perform quality evaluation for each related word (here, a related word candidate) by calculating the average precision AVP using the initial search result.

Further, in the document search method according to the second embodiment of the present invention, a document in which a preset initial search word group appears is ranked searched from a plurality of documents stored in advance,
Step S202 of acquiring the first ranking, and
Step S203 of generating a candidate group of related words related to the initial search word group from the document group included in the ranking of
And a step S204 of performing a ranking search for a document in which a candidate included in the candidate group of the related word appears from the plurality of stored documents, and obtaining a second ranking, and the relation based on the first and second rankings. Since it has steps S205 and S206 for evaluating each candidate included in the word candidate group, it is possible to appropriately perform quality evaluation for each related word (here, related word candidate) as described above. .

In the second embodiment, the case in which the program for implementing the above-mentioned document retrieval method is stored in the memory 3 of the control unit 12 has been described. A computer-readable recording medium (CD-ROM, FD, magneto-optical disk (MO), in which a program for realizing the search method is recorded,
Even if a mini disk (MD) or a rewritable CD (including a CD-RW) is attached to the document retrieval apparatus 100 and the recording medium is read by the CD-ROM drive 8 or the like to execute the program, the same effect is obtained. Is obtained. With this configuration, the program can be easily updated by moving and exchanging the recording medium.

Further, in the second embodiment, the case where the program for realizing the above-mentioned document retrieval method is stored in the memory 3 of the control unit 12 has been described. However, the present invention is not limited to this. The same effect can be obtained by providing the document retrieving apparatus 100 with the above and downloading the program from the external apparatus on the network such as LAN to the control unit 12 and executing the program. With this method, the program can be easily updated via the network.

[0044]

As described above, according to the present invention, the first ranking obtained by using the initial search word group and the second ranking obtained by using the initial search word group and the extended word group are used. By acquiring the new third ranking, a document search device and a document search method having an excellent effect that a pseudo suitability feedback method with less quality deterioration than in the past can be realized even when the quality of a search request is low. And a recording medium can be provided.

Further, the present invention uses the first ranking obtained by using the initial search word group and the related word group (related word candidate group) in the related word search by a method similar to the pseudo compatibility feedback method. A document search device, a document search method, and a record having an excellent effect that the quality of a related word can be appropriately evaluated by evaluating each related word (candidate of the related word) from the second ranking acquired by A medium can be provided.

[Brief description of drawings]

FIG. 1 is a block diagram showing a hardware configuration of a document search device according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing a functional configuration of a document search device according to the first embodiment of the present invention.

FIG. 3 is a flowchart showing a document search method according to the first embodiment of the present invention.

FIG. 4 is a block diagram showing a functional configuration of a document search device according to a second embodiment of the present invention.

FIG. 5 is a flowchart showing a document search method according to the second embodiment of the present invention.

[Explanation of symbols]

1, 2 ranking buffer 3 memory 4 hard disk 5 Input section 6 Output section 7 FDD 8 CD-ROM drive 9 buses 11 Search word buffer (search word storage means) 12 Control unit (ranking acquisition means, related word evaluation means) 13 Search Department 14 Document Storage Unit (Document Storage Unit) 15 Index section 16 Related word buffer 22 CPU 100 Document Retrieval Device (Computer)

Claims (6)

[Claims] 1. A document storage unit that stores a document group to be searched for a ranking, a search word storage unit that stores a first search word group used for a ranking search, and a document storage unit stored in the document storage unit. From the document group, a document in which the first search word group stored in the search word storage means appears is searched to obtain a first ranking, and a second search is made from the document group included in the first ranking. A search word group is generated, and the first search word group and the second search word group are generated from the previously stored document group.
And a ranking obtaining unit that obtains a second ranking by performing a ranking search for a document in which the search word group appears, adjusts the second ranking according to the first ranking, and obtains a third ranking. A document retrieval device characterized by the above. 2. A ranking search is performed for documents in which a preset first search word group appears from a document group stored in advance,
A first step of obtaining a first ranking; a second step of generating a second search term group from a document group included in the first ranking; and a first step from the previously stored document group. The third step of performing a ranking search for documents in which the second search word group and the second search word group appear, and obtaining the second ranking; and the second ranking being adjusted by the first ranking to the third ranking. And a fourth step of obtaining the ranking of the document. 3. A first step of performing a ranking search for a document in which a preset first search word group appears from a document group stored in advance in a computer, and obtaining the first ranking; A second step of generating a second search word group from a document group included in the ranking of 1, and a document in which the first search word group and the second search word group appear from the previously stored document group To perform a third step of searching for the second ranking and obtaining a second ranking, and a fourth step of adjusting the second ranking by the first ranking to obtain the third ranking. A computer-readable recording medium in which a program is recorded. 4. A document storage means for storing a document group to be searched for by ranking, a search word storage means for storing a first search word group used for ranking search, and a document storage means stored in the document storage means. From the document group, a document in which the first search word group stored in the search word storage means appears is searched to obtain a first ranking, and the document group included in the first ranking is used to obtain the first ranking. Generating a related word group related to the search word group, and performing a ranking search for a document in which the related word included in the related word group appears from the document group stored in the document storage unit to obtain a second ranking. , The first
And a related word evaluation unit that evaluates each related word included in the related word group according to the second ranking and the second ranking. 5. A ranking search is performed for documents in which a preset first search word group appears from a document group stored in advance,
A first step of obtaining a first ranking; a second step of generating a related word group related to the first search word group from a document group included in the first ranking; The third step of performing a ranking search for a document in which a related word included in the related word group appears in the related document group to obtain a second ranking, and the related step by the first ranking and the second ranking. A fourth step of evaluating each related word included in the word group. 6. A first step of performing a ranking search for a document in which a preset first search word group appears from a document group stored in advance in a computer, and obtaining the first ranking; A second generating a related word group related to the first search word group from a document group included in the ranking of 1;
And a third step of performing a ranking search for a document in which a related word included in the related word group appears from the previously stored document group, and acquiring a second ranking.
A computer-readable recording medium having recorded thereon a program for executing a fourth step of evaluating each related word included in the related word group according to the first ranking and the second ranking. .