JPH10334106A - Relative word display device and medium where program for relative word display is recorded - Google Patents

Abstract

PROBLEM TO BE SOLVED: To show relative words that can effectively reduce retrieval omissions and narrow down retrieval results. SOLUTION: A retrieval condition accepting means 2 accepts a retrieval condition 'S'. A document retrieving means 3 acquires a document set 'X' meeting the retrieval condition that the retrieval condition accepting means 2 has accepted from a document storage means 1. A relative word calculating means 4 obtains a relative word set 'Wn' by regarding words having high relativity to the retrieval condition as relative words. An appearance frequency calculating means 5 calculates the appearance frequency 'R(Wn)' of each relative word in the relative word set in the document set. A relative word selecting means 6 selects a relative word by using a calculation expression including the number 'N' of documents in the document set and the appearance frequency of each relative word as variables. A relative word display means 7 displays the relative word selected by the relative word selecting means 6.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a related-word presenting apparatus and a medium storing a related-word presenting program. The present invention relates to a medium recording a related word presentation program for presentation.

[0002]

2. Description of the Related Art In a search system for an enormous amount of documents, a search method using a keyword is generally used. When an arbitrary keyword (search term) is input as a search condition into the search system, all documents including the search term in the document content are obtained as search results. The search by this method is called a full-text search. A method is also widely used in which a keyword for search is added to each document in advance, and a document to which a keyword matching the input search word is added is used as a search result.

[0003] A search result obtained by such a search system is a document containing a word completely matching the search entered by the user, or a word completely matching the keyword entered by the user. Is only the document to which is added as a search keyword. Therefore, if an exact match between a search term and a keyword is required,
It cannot be said that all the documents required by the user can be obtained comprehensively.

Therefore, as proposed in Japanese Unexamined Patent Publication No. 2-297290, in order to prevent search omission, a related word of a search word is presented to a user by using a related word dictionary, and the search intention is further improved. A method is used to prompt creation of a matching search formula.

[0005] For example, if the search term entered by the user is "SG
In the case of “ML”, “HTML”, “ODA”, “structured document”, etc. are acquired as related words of “SGML” from the related word dictionary, and presented to the user. The search term is reduced by connecting a related word determined to be appropriate by the user from the presented related words to “SGML” by using a logical sum operator. In this way, by connecting some of the presented related words to the search expression by using a logical sum operator, it is possible to prevent search omission.

[0006] In addition to the purpose of reducing the omission of retrieval, the presented related words can also be used to assist in narrowing down retrieval results. That is, if the number of documents obtained as a search result is too large,
Appropriate refinement is possible by connecting some of the suggested related words with a logical AND operator.

As described above, the user determines whether to prevent search omission or to narrow down search results depending on the number of search results, selects appropriate related words, and connects them to the search formula.

[0008]

By the way, the related words to be connected to the search formula differ between the case where the search is omitted and the case where the search result is narrowed down. Therefore, related words that the user wants to present are also different. In general, when there are few search results, it is desirable to provide extensive and detailed related words in order to prevent search omissions. On the other hand, when there are many search results, the focus is on narrowing the search results,
It is desirable to present related words that can narrow down search results to an appropriate amount, rather than broad or detailed related words.

However, in the above-mentioned prior art, fixed related words are always presented irrespective of the search result, so that it is possible to accurately determine which related words should be connected to the search formula. Can not. That is, it has not been easy to create a search formula that effectively achieves both the purpose of reducing search omissions and narrowing down search results.

The present invention has been made in view of the above points, and provides a related word presentation device that presents related words that can effectively achieve both the purpose of reducing search omissions and narrowing down search results. The purpose is to do.

Another object of the present invention is to provide a related word presentation program for causing a computer to present related words that can effectively achieve both the purpose of reducing search omissions and narrowing down search results. To provide a recorded medium.

[0012]

According to the present invention, in order to solve the above-mentioned problems, in a related word presentation device for presenting a word related to a search condition, a document storage means for storing a plurality of documents, Search condition receiving means for receiving a condition, document search means for obtaining from the document storage means a set of documents that meet the search condition received by the search condition receiving means, and relevance to the search condition received by the search condition receiving means. Related word calculating means for obtaining a set of related words, using words having a high degree as related words, and, in the document set obtained from the document searching means, each related word in the related word set obtained by the related word calculating means. The number of occurrences, which is the number of documents containing the word, is calculated from the number of occurrences calculating unit for each related word, the number of documents in the document set obtained by the document search unit, and the number of occurrences obtained from the number of occurrences calculating unit. A related word selecting means for selecting a related word to be displayed using a calculation formula with the number of occurrences of the related word as a variable, and a related word displaying means for displaying the related word selected by the related word selecting means on a display device And a related word presentation device characterized by having the following.

According to this related word presentation device, when a search condition is input, the search condition is received by the search condition receiving means. Then, the document search means obtains from the document storage means a set of documents that meet the search condition received by the search condition receiving means. Further, the related word calculation means acquires a set of related words using words highly relevant to the search condition received by the search condition receiving means as related words. Then, the number-of-appearance calculation means calculates the number of appearances of each related word in the related word set obtained by the related word calculation means in the document set obtained by the document search means. Then, the related word selection means uses a calculation formula with the number of documents in the document set obtained by the document search means and the number of occurrences of each related word obtained from the number of appearance calculation means as variables, to display the relation to be displayed. The word is selected. The selected related word is displayed on the display device by the related word display means.

Thus, the related words to be displayed can be narrowed down according to the number of documents in the document set obtained by the document search means and the number of appearances of each related word obtained from the appearance number calculation means.

[0015] Further, in a medium storing a related word presentation program for causing a computer to present a word related to a search condition, a document storage means for storing a plurality of documents, a search condition for receiving the input search condition Receiving means, a document search means for obtaining from the document storage means a set of documents that meet the search condition received by the search condition receiving means, and a word highly relevant to the search condition received by the search condition receiving means as a related word A related word calculation means for obtaining a set of related words, in a document set obtained from the document search means, the number of documents containing each related word in the set of related words obtained by the related word calculation means. A certain number of occurrences, the number of occurrences calculating means for each related word, the number of documents in the document set obtained by the document searching means, and the number of occurrences of each related word obtained from the number of occurrences calculating means A computer is used as a related word selecting means for selecting a related word to be displayed using a calculation formula with the current number as a variable, and a related word displaying means for displaying the related word selected by the related word selecting means on a display device. There is provided a medium recording a related word presentation program for causing the program to function.

When the computer executes the related word presentation program recorded on the medium, a document storage means for storing a plurality of documents, a search condition receiving means for receiving input search conditions, and a search condition receiving means A document retrieval unit that acquires a set of documents that match the search condition received from the document storage unit, and a related word that acquires a set of related words using words highly relevant to the search condition received by the search condition reception unit as related words. Word calculation means, occurrence number calculation means for calculating the number of occurrences of each related word in the related word set obtained by the related word calculation means in the document set obtained by the document search means, and document search means obtaining Related word selecting means for selecting related words to be displayed, using a calculation formula in which the number of documents in the set of documents obtained and the number of appearances of each related word obtained from the number of occurrence calculating means are variables, and related word selection A related word display means for displaying the related words stage selected on the display device, but is realized by a computer.

[0017]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a first principle configuration diagram of the present invention. The document storage unit 1 stores a plurality of documents.
The search condition receiving means 2 receives a search condition "S" input by a user using a keyboard or the like. Document search means 3
Obtains from the document storage unit 1 a document set “X” that matches the search condition “S” received by the search condition receiving unit 2. The related word calculation means 4 acquires a related word set “Wn” using words having high relevance to the search condition “S” received by the search condition receiving means 2 as related words. For example, using a related word dictionary, related words of words included in the search condition are extracted and set as a related word set. The number-of-appearance calculation means 5 calculates the number of appearances “R (Wn)” of each related word in the related word set “Wn” obtained from the related word calculation means 4 in the document set acquired by the document search means 3. calculate.

The related word selecting means 6 includes the number of documents “N” of the document set obtained from the document searching means 3 and the number-of-occurrence calculating means 5
The related words to be displayed are selected using a calculation formula using the number of appearances of each related word “R (Wn)” obtained from the above as a variable.
For example, a lower limit value G1 (N), whose value increases stepwise as the number of documents in the document set acquired by the document search means 3 increases, is determined. The upper limit G2 at which the value gradually decreases
(N) is defined. Then, using a calculation formula such that a related word whose number of appearances is equal to or greater than the lower limit and equal to or less than the upper limit is selected,
Select related words to be displayed.

The related word display means 7 displays the related word selected by the related word selecting means 6 on a display device. According to the related word presentation device having such a configuration, related words effective for both purposes of reducing search omission and narrowing down search results are displayed on the display device. That is, when there are many search results N, related words having a large number of appearances in the search results are presented. This becomes an effective related word presentation when the search result is gradually narrowed down.

When the number of search results N is small, related words with a small number of appearances are presented. This results in a wider range of related word presentations that are effective when performing a logical sum connection with a search expression for the purpose of reducing search omissions.

FIG. 2 is a block diagram showing a second principle of the present invention. The major difference between the configuration shown in this figure and the configuration in FIG. 1 is that the processing order of the related word calculation processing and the appearance number calculation processing is reversed. Accordingly, the target of calculation of the number of occurrences is the related word set, which is the word set included in the document set (before the related word calculation processing is executed). However, in the description of this figure, a set of words included in the document set is indicated by the same symbol “Wn” as the related word set.

The document storage means 11 stores a plurality of documents. The search condition receiving means 12 receives the input search condition "S". The document search unit 13 acquires from the document storage unit 11 a document set “X” that matches the search condition received by the search condition receiving unit 12. The number-of-appearance calculating means 14 sets a set of all words existing in the document set obtained by the document searching means 13 as “Wn”, and calculates the number of appearances “R (Wn)” of each word in the document set. .

The related word candidate selecting means 15 calculates the number of documents “N” of the document set obtained by the document searching means 13 and the number of appearances “R (Wn)” of each word obtained from the number of occurrence calculating means 14. Using a calculation formula as a variable, a related word candidate to be displayed is selected.

The related word calculating means 16 extracts related words from related word candidates in the following procedure. First, a first value which is the number of documents obtained by the document search means 13 and a value of each related word candidate which is the number of documents including each related word candidate in the documents obtained by the document search means 13. A second value and a third value for each related word candidate that is the number of documents including each related word candidate among the documents stored in the document storage unit 11 are obtained. Next, a fourth value which is a product or a sum of the first value and the third value is calculated for each related word candidate, and a search condition is calculated based on a ratio of the second value to the fourth value. Receiving means 1
2 calculates the degree of relevance between the received search condition and each related word candidate. Then, related word candidates having a high degree of relevance are extracted as related words. The formula for calculating relevance is:
Extended mutual information, extended Dice-coeff
client and extended t-score can be used.

The related word display means 17 is a related word calculating means 1
The related words extracted by 6 are displayed on the display device. According to such a related-word presentation device, related-word candidates are selected before calculating the related-degree. Therefore, it is possible to omit the related-degree calculating process for related words that do not require the related-degree calculation. It becomes possible. Therefore, even when performing a complicated relevance calculation, it is possible to display related words effective for narrowing down documents at high speed.

FIG. 3 is a block diagram showing the third principle of the present invention. This related word presentation device includes a document storage unit 21, a search condition receiving unit 22, a document search unit 23, and an appearance number calculation unit 2.
4. Related word candidate selecting means 25, related word calculating means 26, related word pair appearance number calculating means 27, related word related degree calculating means 2
8 and related word display means 29. Here, the document storage means 21, the search condition receiving means 22, the document search means 23, the number-of-appearance calculation means 24, the related word candidate selection means 25,
The related word calculating means 26 has the same function as the component of the same name shown in FIG. 2 and therefore will not be described, and the other components will be described below.

When an arbitrary related word is specified, the related word pair appearance number calculating means 27 determines the specified specific related word and another related word in the related word set obtained from the related word calculating means 26. Is calculated for the document set acquired by the document search means 23.

The related word interrelationship calculating means 28 calculates the number of occurrences of each related word obtained from the number of occurrence calculating means 24, the specific related word obtained from the related word pair appearance number calculating means 27, and other related words. Then, the related word relevance is calculated based on a calculation formula using the number of simultaneous appearances between and as variables, and related words having a high related word relevance with respect to the specific related word are extracted.

The related word display means 29 is a related word calculating means 2
6 displays the related words extracted on the display device, and also displays the related words extracted by the related word relevance calculating means 28 on the display device.

With such a related word presentation device, a user can specify another related word and know another related word having a certain relationship with the related word. The related words to be connected by the child can be easily selected.

The configuration shown in FIG. 3 is obtained by adding related word pair occurrence number calculating means and related word interrelation degree calculating means to the structure shown in FIG. 2 and adding a function of related word display means. ,
Similar means and functions can be added to the configuration of FIG.

The functions of the constituent elements of the above-described respective principle configurations can be realized by executing a program in which instructions of the respective processing functions are described by a computer. In this case, the program is stored in a computer-readable recording medium. As a recording medium, a semiconductor storage device,
A magnetic recording device, an optical disk, or the like can be used.

By the way, in the related word calculating means of the present invention,
Mutual information, Dice-coefficient and t-sc, which are statistic values originally used as similarities between words
By expanding ore, the similarity between a search expression and a word can be calculated, and a word having a high similarity can be used as a related word. Mutual information, Dice-coefficien
Examples of using t and t-score to calculate similarity between words include “Haruno, Yamazaki: Bilingual Alignment Using Dictionaries and Statistics, IPSJ SIG Technical Report,
96-NL-112, pp. 23-30 (199
6) ”,“ Omori, Tsutsumi, Nakanishi: Creating a bilingual word dictionary using statistical information, Proceedings of the 2nd Annual Meeting of the Association for Language Processing,
pp. 49-52 (1996) ".

An extension method for applying the mutual information and the like to the present invention will be described below. Word word1
The mutual information (MI) between and word2 is

[0035]

(Equation 1)

Is defined as Here, the number of documents to be searched is M, the number of documents including both word1 and word2 is a,
If the number of documents containing only word1 is b and the number of documents containing only word2 is c,

[0037]

(Equation 2)

[0038]

(Equation 3)

[0039]

(Equation 4)

Is as follows. On the other hand, in the present invention, the mutual information (MI ₀ ) between the search expression S and the word word is calculated as

[0041]

(Equation 5)

Is defined as However, the total number of documents to be searched includes M and word, and the number of documents obtained from the search formula S includes a ₀ , and the number of documents obtained from the search formula S that do not include the word includes b ₀ and word. When the number of documents excluding the documents obtained from the search formula S is c ₀ ,

[0043]

(Equation 6)

[0044]

(Equation 7)

[0045]

(Equation 8)

Is as follows. Here, “a ₀ + b ₀ ” corresponds to “first value” in the description of FIG. 2, and “a ₀ ” corresponds to “second value”.
And “a ₀ + c ₀ ” corresponds to the “third value”. Therefore, equation (5) becomes

[0047]

(Equation 9)

Thus, the total number of documents to be searched M,
It is a calculation formula using “first value”, “second value”, and “third value” as variables. Dice-coefficie is a statistic for calculating the similarity between words in the same manner as the mutual information.
nt and t-score can be mentioned. Di
ce-coefficient (DC) and t-sc
ore (TS)

[0049]

(Equation 10)

[0050]

[Equation 11]

Is defined as For these, similarly to the mutual information amount, the following extension can be applied in order to calculate the similarity between the search formula and the word.

[0052]

(Equation 12)

[0053]

(Equation 13)

MI ₀ (S, word), DC ₀ (S, wor
Both d) and TS ₀ (S, word) mean that the greater the value, the higher the similarity between the search formula S and the word word. Thereafter, MI ₀ (S, word) is “extended mutual information”, DC ₀ (S, word) is “extended DC”, TS ₀
(S, word) is called an “extended TS”. In addition,
Like the mutual information, the extended DC and the extended TS are respectively

[0055]

[Equation 14]

[0056]

(Equation 15)

Can be expressed as follows. As can be seen from Expression (14), when obtaining the extended DC, the total number M of documents to be searched is unnecessary. Next, an embodiment of the related word presentation device of the present invention will be specifically described.

FIG. 4 is a block diagram showing the configuration of the embodiment of the present invention. This is an embodiment based on the configuration shown in FIG. The document storage unit 31 is a storage device that stores the contents of the digitized search target document in pair with the document identifier added by the morphological analysis unit 32.

The morphological analysis means 32 includes the document storage means 31
After adding a document identifier to each document stored in
Each document is subjected to morphological analysis processing to extract independent words (words to be keywords) and store them in pairs with corresponding document identifiers.

The index structure generating means 33 generates an index structure based on the result of the morphological analysis processing by the morphological analyzing means 32.
Word-word identifier list 34a, word identifier-document identifier list 34b, document identifier-word identifier list 34c
Create

The index structure storing means 34 stores the word-word identifier list 34 created by the index structure generating means 33.
a, a storage device for storing a word identifier-document identifier list 34b and a document identifier-word identifier list 34c.

The word-word identifier list 34a is a list describing the correspondence between word character strings and word identifiers for specifying the word. The word identifier-document identifier list 34b is a list that describes, for each word identifier, a set of document identifiers of documents that include the word character string indicated by the word identifier.

The document identifier-word identifier list 34c is
For each document identifier, a list describing a set of word identifiers of words included in the document indicated by the document identifier.

The search condition receiving means 41 is a user interface for receiving, from an input device such as a keyboard, an input of a search condition (search expression) formed by connecting words by a logical sum operator or a logical product operator. .

The document search means 42 is a search condition receiving means 4
The document identifiers of all the documents that match the search condition input in 1 are acquired with reference to the word-word identifier list 34a and the word identifier-document identifier list 34b, and the acquired document identifier set is stored. Further, the number of identifiers in the stored document identifier set is passed to the related word candidate selecting means 45. To the related word calculating means 46, the number of identifiers in the stored document identifier set and the total number of documents including the word corresponding to the word identifier given from the related word calculating means 46 are passed. The total number of documents that include the word identifier pair given from the related word pair appearance number calculation means 47 and are included in the stored document identifier set is passed to the related word pair appearance number calculation means 47.

The in-document word search means 43 obtains an identifier set of words contained in each document of the document set meeting the search condition obtained from the document search means 42 with reference to the document identifier-word identifier list 34c. , Connect them and 1
One word identifier set.

The word appearance number calculation means 44 calculates the number of occurrences of each word identifier in the set of word identifiers obtained from the word search means 43 in the document, and creates a list by pairing the word identifiers and the number of occurrences. I do.

The related word candidate selection means 45 calculates the number of occurrences corresponding to each word identifier obtained from the word occurrence number calculation means 44 and the total number of document identifiers which meet the search conditions obtained from the document search means 42. , Calculate word identifiers that are candidates for related words.

The related word calculating means 46 calculates the number of occurrences corresponding to each word identifier which is a candidate of the related word calculated by the related word candidate selecting means 45 and the document which matches the search condition obtained from the document searching means 42. Based on three values of the total number of identifiers and the number of documents including the word corresponding to the word identifier obtained from the document search means 42, the search condition input to the search condition receiving means 41 and the related word candidate selecting means 45 The extended mutual information between each of the calculated related word candidate words is calculated. Then, a related word candidate word having an expanded mutual information amount larger than a predetermined threshold value is set as a related word.

The related word pair appearance number calculating means 47 obtains the number of occurrences of each related word from the related word calculating means 46, and further obtains a document set matching the search condition input to the search condition receiving means 41. , The number of documents containing any two related word pairs at the same time is acquired from the document search means 42.

The related word relevance calculating means 48 calculates the number of occurrences of each related word and the number of documents including each related word pair in a document set which meets the search condition obtained from the related word pair occurrence number calculating means 47. Based on this, the degree of relevance (relevance) between related words is calculated.

The related word display means 49 is provided for the related word calculating means 4.
Each related word calculated in step 6 is converted into a related word relevance calculating means 4
8 is a user interface that outputs in accordance with the related word relevance calculated by step 8.

The search result display means 50 is provided for the document search means 4
This is a user interface that outputs a document set that matches the search condition input to the search condition receiving unit 41 obtained from the document storage unit 2 with reference to the document storage unit 31.

The functions of the above-described components are realized by a computer executing a predetermined program module. A computer program for realizing the above is recorded on a recording medium such as a semiconductor memory or a magnetic recording medium. However, the document storage means 31 and the index structure storage means 34
Is a function realized by controlling a storage device such as an actual HDD (hard disk device).

Each component of the related word presentation device of FIG. 4 has the following relationship with the component of FIG. The document storage means 31 and the index structure storage means 34 correspond to the document storage means 21. The search condition receiving means 41 corresponds to the search condition receiving means 22. The document search means 42 corresponds to the document search means 23. The in-document word search means 43 and the word appearance number calculation means 44 correspond to the appearance number calculation means 24. The related word candidate selecting means 45 corresponds to the related word candidate selecting means 25. The related word calculating means 46 corresponds to the related word calculating means 26. The related word pair appearance number calculation means 47 corresponds to the related word pair appearance number calculation means 27. The related word relevance calculating means 48 corresponds to the related word relevance calculating means 28. The related word display means 49 corresponds to the related word display means 29.

Here, in this embodiment, it is necessary to execute an index structure generating process before searching for a related document. Therefore, first, a process of generating an index structure will be described.

As a precondition for the index structure generation processing, a morphological analysis result list must be generated. FIG.
FIG. 4 is a diagram showing an example of a morphological analysis result list 32a stored in the morphological analysis means 32. Morphological analysis means 32
Assigns an identifier to each document to be searched stored in the document storage means 31, performs morphological analysis processing on each document, extracts an independent word, and stores the independent word as a pair with the corresponding document identifier. However, when the same independent word is extracted a plurality of times from the same document, the second and subsequent extraction results are ignored, and the independent word corresponding to one document identifier does not overlap.

Based on the morphological analysis result list 32a,
The index structure generating means 33 generates various index structures. FIG.
8 to 8 show examples of index structures created by the index structure generating means 33 and stored in the index structure storing means 34. FIG. Note that the data in FIGS. 6 to 8 is an example created based on the data in FIG.

FIG. 6 is a diagram showing an example of a word-word identifier list. The word-word identifier list 34a stores a pair of the extracted word and the identifier assigned to the word.

FIG. 7 is a diagram showing an example of a word identifier-document identifier list. Word identifier-document identifier list 34
In b, a word identifier and an identifier (document identifier) of a document including the word to which the word identifier is assigned are stored as a set.

FIG. 8 is a diagram showing an example of a document identifier-word identifier list. Document identifier-word identifier list 34
In c, a document identifier and a word identifier of a word included in the document to which the document identifier is assigned are stored as a set.

The algorithm for generating the index structure by the index structure generating means 33 is as follows. FIG. 9 is a flowchart showing an algorithm for generating an index structure. [S1] Generation of Word-Word Identifier List 34a A list is created in which all words in the morphological analysis result list stored in the morphological analysis unit 32 are sorted without duplication and in the order of the values of the word character strings. . For each word, a natural number starting with 1 from the top of the list is assigned as a word identifier. [S2] Generation processing of the document identifier-word identifier list 34c Each word in the morphological analysis result list stored in the morphological analysis unit 32 is replaced with the word identifier assigned in step S1, and a corresponding word identifier for each document identifier Are sorted in ascending order. [S3] Generation process of the word identifier-document identifier list 34b The word identifiers are arranged in order from 1, and the document identifier of the document including the word corresponding to each word identifier is determined by the document identifier-word identifier list 34c created in step S2. It extracts by reference, and stores it as a pair with a word identifier.

An index structure is generated by the above algorithm. After the generation of the index structure is performed, the search condition receiving means 41 can input a search formula. Then, when the user inputs a desired search formula using an input device such as a keyboard and issues a search start instruction, the related word presentation process is started.

FIGS. 10 and 11 show search condition receiving means 41.
2 shows an algorithm for obtaining a related document from the search formula input in FIG. FIG. 10 is the first half of a flowchart showing the processing procedure of the present invention, and FIG. 11 is the second half of the flowchart showing the processing procedure of the present invention. Hereinafter, each step of FIGS. 10 and 11 will be described. In the following description, the word-word identifier list 34a is L1, the word identifier-document identifier list 34b is L2, and the document identifier-
The word identifier list 34c is described as L3. [S11] The search condition receiving means 41 receives a search expression in which words are combined by a logical AND operator or a logical OR operator.
This search formula is called S. [S12] The document search means 42 acquires the document identifier of the document conforming to S with reference to L1 and L2. The obtained document identifier set is called X, and the number of elements of the set X is N. [S13] If N = 0 in step S12, the process proceeds to step S14; otherwise, the process proceeds to step S15. [S14] The related word display means 49 displays that there is no related document of S, and ends. [S15] The intra-document word search means 43 acquires all word identifiers corresponding to each document identifier belonging to X with reference to L3. A set of acquired word identifiers is Y. [S16] The word appearance number calculation means 44 removes duplication of word identifiers belonging to Y and records the number of duplications of each word identifier. A new word identifier set with duplicates removed
, And the number of duplications of the element Wn of Y (n = 1, 2,..., P) is R (Wn). Here, P is the number of elements of Y. [S17] The related word candidate selection unit 45 determines that G1 (N) ≦ R
Wn that does not satisfy (Wn) ≦ G2 (N) is removed from the elements of Y, and the obtained set is newly defined as Y, and the element Wn of Y
The number of duplications of (n = 1, 2,..., P) is R (Wn). Here, P is newly set as the number of elements of Y. Where G
1 (x) is an increasing (step) function for x, G2 (x)
Is the decreasing (step) function for x.

For example, G1 (x) and G2 (x) can be functions as follows.

[0086]

G1 (x) = [(20 × n−19 ≦ x ≦ 20)
× n), a natural number satisfying n)

[0087]

[Equation 17]

[S18] The document search means 42 acquires the total number of document identifiers corresponding to Wn for all word identifiers Wn (n = 1, 2,..., P) belonging to Y from L2. Then, the number of document identifiers corresponding to the element Wn of Y is
(Wn). [S19] The related word calculation unit 46 sets the number of all search target documents to M for the word identifier Wn (n = 1, 2,..., P) belonging to Y,

[0089]

(Equation 18)

[0090]

[Mathematical formula-see original document] prob (Wn) = F (Wn) / M (19) is calculated, and these values are combined with Wn to form a list.
Also,

[0091]

Calculate prob (S) = N / M (20) [S20] The related word calculating means 46 calculates the extended mutual information MI ₀ (S, Wn) for each of the word identifiers Wn (n = 1, 2,..., P) belonging to Y in accordance with equation (5). calculate. [S21] With respect to a preset threshold T, T ≦ MI ₀
Wn satisfying (S, Wn) (n = 1, 2, 3,...,
If P) exists, the process proceeds to step S22; otherwise, the process proceeds to step S23. [S22] The related word display means 49 determines that T ≦ MI ₀ (S, W
n), the words corresponding to Wn (n = 1, 2, 3,..., P) are output as related words of S in ascending order of the value of MI ₀ (S, Wn), and the process ends. [S23] The related word display means 49 displays that there is no related document of S, and ends.

With the above processing, it is possible to obtain a related word from the retrieval formula. In addition, in step S17, if there are many documents (search results) that match the search formula S, the related words are calculated based on words having a large number of occurrences in the search results, and the search results are gradually narrowed down. In this case, it is possible to present effective related words. It should be noted that a word having an extremely large number of appearances is a common word and is excluded as inappropriate for presentation as a related word. Step S1
G2 in the equation (7) is a function for removing such an extremely high-frequency word.

When the number of search results is small, related words are calculated by including words having a small number of occurrences.
It is possible to present, in a wider range, related words for performing a logical sum connection with a search expression for the purpose of reducing search omission.

The algorithm for calculating the related word relevance (relation between related words) by the related word pair appearance number calculating means 47 and the related word relevance calculating means 48 is as follows. FIG.
2 is a flowchart showing a procedure for calculating the degree of association between related words. [S31] Let Z be the related word set of S obtained by the above processing of steps S11 to S23, and let V be the element of Z
(M =, 2, 3,... Q). Here, Q is the number of elements of Z. [S32] The related word pair appearance number calculating means 47 calculates an arbitrary two related word pairs (Vm1, Vm2) (m1 = 1, 2,...,
Q, m2 = 1, 2,..., Q, m1 ≠ m2),
a document containing both m1 and Vm2 at the same time and step S
Then, the total number of documents (= R2 (Vm1, Vm2)) belonging to the document set corresponding to X obtained in step 12 is acquired. [S33] The related word relevance calculating means 48 calculates Vm (m = m
1, 2,..., Q), upper (Vm), lo
lower (Vm), equivalent (Vm), si
Calculate milar (Vm). Here, upper (V
m), lower (Vm), equivalent (V
m) and similar (Vm) are subsets of Z whose elements are Vn (n = 1, 2,..., Q) satisfying the following conditions. upper (Vm): Tu1 ≦ R (Vn) / R (V
m), Tu2 ≦ R2 (Vm, Vn) / R (Vm) lower (Vm): T11 ≦ R (Vm) / R (V
n), Tl2 ≦ R2 (Vm, Vn) / R (Vm) equivalent (Vm): Tr1 ≦ R (Vn) /
R (Vm) ≦ Tr2, R2 (Vm, Vn) / R (Vm)
≦ Tr3 similar (Vm): Ts1 ≦ R (Vn) / R (V
m) ≦ Ts2, Ts3 ≦ R2 (Vm, Vn) / R (V
m) where Tu1, Tu2 (≦ 1), T11, T12 (≦
1), Tr1, Tr2 (≧ Tr1), Tr3, Ts1,
Ts2 (≧ Ts1) and Ts3 are preset constants.

For example, the following constants are set. Tu1 = Tl1 = 4 Tu2 = Tl2 = Tr1 = Ts1 = Ts3 = 0.9 Tr2 = Ts2 = 1.1 Tr3 = 0.1 The above expressions have the following meanings.

The upper (Vm) is included in a larger number of documents than the document containing Vm, and is a related word contained in most of the documents containing Vm. lower (V
m) is included in only a smaller number of documents than the document containing Vm, and V
It is a related word that also includes m.

Equivalent (Vm) is a related word that is included in the same number of documents as Vm and hardly included in the same document as Vm.
Similar (Vm) is a related word that is included in the same number of documents as Vm and is included in almost all documents including Vm.

FIG. 13 is a diagram conceptually showing the relationship between related words Vn and Vm. Note that each document set is all included in a document set X that meets the search formula S. (A) is e
The relationship between the document set 61 including the related word Vn included in the qualient (Vm) and the document set 71 including the related word Vm is shown. As shown in this figure, the document set 61 including the related word Vn and the document set 71 including the related word Vm have almost the same amount of documents. The number of documents included in both document sets is very small (or not at all).

(B) shows the relationship between a document set 62 containing related words Vn included in similar (Vm) and a document set 72 containing related words Vm. As shown in this figure, the document set 62 including the related word Vn and the document set 72 including the related word Vm have almost the same amount of documents. Most of the documents containing one related word include the other related word.

(C) shows the relationship between the document set 63 including the related word Vn included in the upper (Vm) and the document set 73 including the related word Vm. As shown in this figure, most of the document set 73 including the related word Vm is associated with the related word Vn.
Are included in the document set 63 including In addition, the document set 63 including the related word Vn has a larger number of documents.

(D) shows the relationship between the document set 64 including the related word Vn included in the lower (Vm) and the document set 74 including the related word Vm. As shown in this figure, most of the document set 64 including the related word Vn is associated with the related word Vm.
Is included in the document set 74 including Moreover, the document set 74 including the related word Vn has a larger number of documents.

If these relationships are presented to the user together with related words, the user can more appropriately select related words for narrowing down when there are many search results. . Next, a user interface in a case where the related words classified as described above are presented to the user will be described.

FIG. 14 is a diagram showing a related word search screen. The related word search screen 80 is divided into three sub-windows 81 to 83. The subwindow 81 is a window for inputting a search expression, and is provided with a text input field 81a and a search button 81b. The user inputs a search expression into the text input field 81a using an input device such as a keyboard,
By pressing 1b, a search command can be input. The sub window 81 provides a function corresponding to the search condition receiving means 41.

The sub window 82 has a related word display field 82a and a related word relation display field 82b. In the related word display field 82a, related words related to the search formula input in the text input field 81a are displayed. Related word relation display field 82
In b, a related word having a predetermined relationship with the selected related word in the related word display field 82a is displayed.

The sub window 83 is a window for displaying a search result. In this sub-window 83, document information matching the search formula entered in the text input field 81a is displayed.

A user who searches for a document using such a user interface first inputs a search expression into the text input field 81a using an input device such as a keyboard, and presses a search button 81b. Then
The search formula is received by the related document search device by the search condition receiving means 41. Then, steps S11 to S23
Is performed. The related word obtained as a result of the processing is displayed in the related word display field 82 in the sub window 82.
a. Further, in the sub-window 83, the document information of the document set corresponding to the document identifier set X obtained in step S12 is displayed.

By selecting one of the related words displayed in the related word field 82a, steps S21 to S2 are performed.
Step 3 is performed. As a result, the specified related word u
upper (Vm), lower (Vm), equiva
Related words belonging to lent (Vm) and similar (Vm) are obtained. These related words are displayed in the related word relation display field 82b.

FIG. 15 is a diagram showing a display example when "airplane" is input as a search formula. In the text input field 81a, the input search expression "airplane" is displayed. In the related word display field 82a, related words obtained from "airplane" are displayed. In the sub-window 83, document information of a document including "airplane" is displayed.

In this example, since there are many search results, the related word calculation is performed by the process of step S17 with the words having a high number of appearances in the search results as related word candidates. Therefore, even when the search is performed using the displayed related words (when the search is connected to the original search expression using a logical product operator), the search is not performed excessively, and an excessive increase in search omission can be prevented.

FIG. 16 is a diagram showing a display example when the related word “main wing” is specified. When the user designates “main wing” in the related word display field 82a with a mouse cursor, a related relation (upper, lower, equi) with “main wing” is displayed in the related word relation display field 82b.
related words having the value (valent, similar) are displayed. The user creates a search formula for narrowing down while referring to these relationships. Thereby, an appropriate search formula can be created.

FIG. 17 is a diagram showing a display example when "(fighter and cockpit and missile)" is input as a search expression. In the text input field 81a, the entered search expression "(fighter and cockpit a
nd missile) "is displayed. In the related term display field 82a, "(fighter and cockpit and
Missile) "is displayed. The sub window 83 displays document information of a document including all of “fighters”, “cockpits”, and “missiles”.

In this case, since the search result is small, the related word calculation is performed by the processing in step S17 including the word having a low appearance frequency in the search result. Therefore, detailed related words are displayed, and noise (a document that does not match the purpose of the search) can be generated by adding related words that the user thinks necessary to the search expression (connecting with a logical sum operator).
It is possible to obtain a search result with less number.

[0113]

As described above, in the related word presentation device of the present invention, the number of documents in the document set obtained by the document search means and the number of appearances of each related word obtained from the number of appearance calculation means are used as variables. Because the related words to be displayed are selected by using the calculation formula, some documents that match the search conditions,
The displayed related word can be changed according to the appearance frequency of the related word. As a result, it is possible to selectively present relevant terms that are effective in narrowing down search results,
Both objectives of reducing search omissions and narrowing search results can be effectively achieved.

Further, the medium storing the related-word presenting program of the present invention executes the recorded related-word presenting program on a computer. The related words can be selectively displayed by using a calculation formula that uses, as variables, the number of occurrences of each related word obtained from the appearance number calculation means. As a result, it is possible to cause the computer to present relevant words effective for narrowing down the search results.

[Brief description of the drawings]

FIG. 1 is a first principle configuration diagram of the present invention.

FIG. 2 is a second principle configuration diagram of the present invention.

FIG. 3 is a third principle configuration diagram of the present invention.

FIG. 4 is a block diagram showing a configuration of an embodiment of the present invention.

FIG. 5 is a diagram showing an example of a morphological analysis result list stored in a morphological analysis unit.

FIG. 6 is a diagram showing an example of a word-word identifier list.

FIG. 7 is a diagram illustrating an example of a word identifier-document identifier list.

FIG. 8 is a diagram illustrating an example of a document identifier-word identifier list.

FIG. 9 is a flowchart illustrating an algorithm for generating an index structure.

FIG. 10 is the first half of a flowchart showing the processing procedure of the present invention.

FIG. 11 is the second half of a flowchart showing the processing procedure of the present invention.

FIG. 12 is a flowchart showing a procedure for calculating the degree of association between related words.

FIG. 13 is a diagram conceptually showing a relationship between related words Vn and Vm. (A) shows a relationship between a document set including the related word Vn included in the equivalent (Vm) and a document set including the related word Vm. (B) is similar
The relationship between the document set including the related word Vn included in (Vm) and the document set including the related word Vm is shown. (C) is u
The relationship between the document set including the related word Vn and the document set including the related word Vm included in the pupper (Vm) is shown.
(D) shows a relationship between a document set including the related word Vn included in lower (Vm) and a document set including the related word Vm.

FIG. 14 is a diagram showing a related word search screen.

FIG. 15 is a diagram illustrating a display example when “airplane” is input as a search expression.

FIG. 16 is a diagram illustrating a display example when a related term “main wing” is designated.

FIG. 17 is a diagram showing a display example when “(fighter and cockpit and missile)” is input as a search expression.

[Explanation of symbols]

 1 Document storage means 2 Search condition receiving means 3 Document search means 4 Related word calculation means 5 Appearance count calculation means 6 Related word selection means 7 Related word display means

Claims (13)

[Claims] 1. A related word presentation device for presenting a word related to a search condition, comprising: a document storage unit for storing a plurality of documents; a search condition receiving unit for receiving input search conditions; A document search unit that obtains a document set that meets the received search condition from the document storage unit; and a related word set that obtains a related word set using words highly relevant to the search condition received by the search condition reception unit as related words. Word calculation means, in a document set obtained from the document search means, the number of occurrences which is the number of documents containing each related word in the related word set obtained by the related word calculation means, Number of occurrences calculating means to calculate for each, the number of documents of the document set obtained by the document search means, using a calculation formula with the number of occurrences of each related word obtained from the number of occurrences calculation means as a variable, display A related word presentation device, comprising: related word selection means for selecting a related word to be used; and related word display means for displaying the related word selected by the related word selection means on a display device. 2. The related word selecting means determines a lower limit value at which the value increases stepwise as the number of documents in the document set acquired by the document search means increases. The related word presentation device according to claim 1, wherein the related word to be displayed is selected using a calculation formula to be selected. 3. The related word selecting means determines an upper limit value whose value decreases stepwise as the number of documents in the document set acquired by the document search means increases. The related word presentation device according to claim 1, wherein the related word to be displayed is selected using a calculation formula to be selected. 4. When an arbitrary related word is specified, the number of simultaneous occurrences of the specified specific related word and other related words in the related word set obtained from the related word calculating means is calculated by the document A related word pair occurrence count calculating unit that calculates the document set acquired by the search unit as a target, and an appearance count of each related word obtained from the occurrence count calculation unit,
Calculate the related word interrelationship based on a calculation formula with the number of simultaneous occurrences between the specific related word and the other related words obtained from the related word pair appearance number calculating means as a variable, and And a related word relevance calculating means for extracting a related word having a high related word relevance between the related words. The related degree displaying means displays a related word extracted by the related word calculating means on a display device. The related word presentation device according to claim 1, wherein the related word extracted by the related word relevance calculating means is displayed on the display device while being displayed. 5. A related word presentation device for presenting a word related to a search condition, a document storage unit for storing a plurality of documents, a search condition receiving unit for receiving input search conditions, and the search condition receiving unit. A document search unit that obtains a document set that meets the received search condition from the document storage unit; and the number of occurrences that is the number of documents containing each word present in the document set obtained by the document search unit, Using an appearance number calculation unit that calculates for each word, a number of documents of the document set obtained by the document search unit, and a calculation formula using the number of appearances of each word obtained from the occurrence number calculation unit as variables, A related word candidate selecting unit that selects a related word candidate; a first value that is the number of documents obtained by the document searching unit; and a second value that is the number of occurrences of each related word candidate obtained by the occurrence number calculating unit. And the value of A third value for each related word candidate, which is the number of documents including each related word candidate, among the documents stored in the document storage means, is obtained, and the first value and the third value are obtained. A fourth value, which is a product or a sum of the above, is calculated for each related word candidate, and the search condition received by the search condition receiving means and each related word Related word calculation means for calculating the degree of relevance between candidates and extracting a related word candidate having a high degree of relevance as a related word, and related word display means for displaying the related word extracted by the related word calculation means on a display device And a related word presentation device, characterized by having: 6. The related word candidate selecting means determines a lower limit value at which the value increases stepwise as the number of documents in the document set obtained by the document searching means increases, and the related word whose appearance number is equal to or more than the lower limit value The related-word presentation device according to claim 5, wherein the related-word to be displayed is selected using a calculation formula for selecting the related-word. 7. The related word candidate selecting means determines an upper limit value whose value decreases stepwise as the number of documents in the document set obtained by the document searching means increases, and a related word whose number of appearances is equal to or less than the upper limit value. The related-word presentation device according to claim 5, wherein the related-word to be displayed is selected using a calculation formula for selecting the related-word. 8. The related word calculation means sets the number of all documents stored in the document storage means to M, a first value to α, a second value to each related word candidate to β, When the third value for each word candidate is γ, the value of the extended mutual information obtained by the following formula, extended mutual information = log ₂ {(Mβ) / (αγ)}, is used as the search condition. 6. The related-word presenting apparatus according to claim 5, wherein the degree of relevance between the search condition received by the receiving unit and each related-word candidate is set. 9. The related word calculation means sets the number of all documents stored in the document storage means to M, sets a first value to α, sets a second value for each related word candidate to β, When the third value for each word candidate is γ, the search condition receiving means obtains the value of the extended TS obtained by the following equation: extended TS (t-score) = M {(Mβ−αγ) / (αγ)} 6. The related-word presentation device according to claim 5, wherein the degree of relevance is between the received search condition and each related-word candidate. 10. The related word calculating means calculates the following value when the first value is α, the second value for each related word candidate is β, and the third value for each related word candidate is γ: The extended DC (Dice-coefficent) = 2β / (α + γ) value is used as the degree of association between the search condition received by the search condition receiving means and each related word candidate. The related word presentation device according to claim 5, wherein 11. When an arbitrary related word is specified, the number of simultaneous occurrences of the specified specific related word and another related word in the related word set obtained from the related word calculating means is calculated by the document A related word pair occurrence count calculating unit that calculates the document set acquired by the search unit as a target, and an appearance count of each related word obtained from the occurrence count calculation unit,
Calculate the related word interrelationship based on a calculation formula with the number of simultaneous occurrences between the specific related word and the other related words obtained from the related word pair appearance number calculating means as a variable, and And a related word relevance calculating means for extracting a related word having a high degree of related word relevance between the related words. The related word presentation device according to claim 5, wherein the related word extracted by the related word relevance calculation means is also displayed on the display device. 12. A medium storing a related word presentation program for causing a computer to present a word related to a search condition, a document storage means for storing a plurality of documents, a search condition for receiving the inputted search condition. Receiving means; a document searching means for obtaining, from the document storing means, a set of documents which match the search conditions received by the search condition receiving means; words having high relevance to the search conditions received by the search condition receiving means as related words A related word calculation means for obtaining a set of related words, in the document set obtained from the document search means, the number of documents containing each related word in the set of related words obtained by the related word calculation means An occurrence count calculating means for calculating a certain number of occurrences for each related word, the number of documents in a document set obtained by the document search means, and A computer as a related word selecting means for selecting a related word to be displayed by using a calculation formula with the number of appearances as a variable, and a related word displaying means for displaying the related word selected by the related word selecting means on a display device. A medium on which a related word presentation program for functioning is recorded. 13. A medium storing a related word presentation program for causing a computer to present a word related to a search condition, a document storage means for storing a plurality of documents, a search condition for receiving the inputted search condition. Receiving means; a document searching means for obtaining, from the document storing means, a document set meeting the search condition received by the search condition receiving means; a document containing each word present in the document set obtained by the document searching means. The number of appearances, which is the number of occurrences, is calculated for each word, the number of documents in the document set obtained by the document search means, and the number of occurrences of each word obtained from the number of occurrences are variable. A related word candidate selecting unit that selects a related word candidate, a first value that is the number of documents acquired by the document search unit, and an occurrence number calculation unit that acquires A second value, which is the number of occurrences of each related word candidate, and a third value, which is the number of documents containing each related word candidate, among the documents stored in the document storage means. Is calculated, and a fourth value which is a product or a sum of the first value and the third value is calculated for each related word candidate, and based on a ratio between the second value and the fourth value. Related word calculating means for calculating the degree of relevance between the search condition received by the search condition receiving means and each related word candidate, and extracting related word candidates having a high degree of relevance as related words, And a related-word displaying means for displaying the related-words extracted by the computer on a display device, and a medium storing a related-word presenting program for causing a computer to function.