JP2000227917A - Thesaurus browsing system and method therefor and recording medium recording its processing program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the use efficiency of a related thesaurus by permitting a user to select any set, extracting a term related with a general term belonging to the respective representative terms of the selected set from the thesaurus and generating and displaying several sets from the extracted terms. SOLUTION: The set of terms in document data stored in a document data storage part 2a is stored in a cooccurrence term table 3a. The term extracted from a document and the frequency are stored in the document term table 3b for every document. A related thesaurus is generated from document data in a related thesaurus generation part 1a. The term vectors of the respective documents are extracted from a term vector extraction part 1b and the view of the related thesaurus is generated. A term cluster list is selected from it and the related term of the term belonging to the list is extracted. Several sets are generated from the extracted terms and they are displayed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for displaying terms stored in a thesaurus, and more particularly, to a thesaurus browsing system suitable for improving the efficiency of searching terms stored in a network type thesaurus. The present invention relates to a method and a recording medium on which a processing program is recorded.

[0002]

2. Description of the Related Art First, the outline of this thesaurus will be briefly described with reference to JIS standards, and then the features of a thesaurus storing terms (related terms) targeted by the present invention will be described. 1 shows a conventional technique for browsing the thesaurus.

According to JIS X 0901 "Thesaurus construction and method of creation", a thesaurus is a "vocabulary of a controlled index language, and organized in such a way that a priori relationship between concepts is explicitly specified in advance. What did you do? " Here, the “vocabulary of the index language”, that is, the “index word” is “an appropriate expression of a concept in the form of a noun or a phrase equivalent to a noun” (JIS X0901), and is generally called “term”. Often called. Therefore, it is referred to as “term” below.

There are "synonymous relationship", "hierarchical relationship", and "association relationship" as the relationship between terms described in the thesaurus (JIS X 0901). The “synonymous relationship” is a relationship between terms indicating the same concept. For example, "camera" and "camera". A “hierarchical relationship” is a relationship between a superordinate concept and a subordinate concept, such as typological relationships (relationships between classes and their members, eg, “primates” and “monkeys”), and whole partial relationships (part names and their names). There are three types of names, such as "digestive organs" and "gut", and example relationships (relationship between categories and their cases, such as "high-speed rail" and "Shinkansen"). A thesaurus storing the terms having such a hierarchical relationship is hereinafter referred to as a “hierarchical thesaurus”.

[0005] On the other hand, "relationship" means "relationship between words having relations other than synonymous relation and hierarchical relation" (JIS X
0901), meanings partially overlap (eg, “passenger car” and “private car”), or when one term strongly implies the other (eg, “publishing” and “books”). )). As a feature of a term having such a related relationship, that is, a thesaurus storing the related terms (hereinafter, referred to as a “related thesaurus”), it is pointed out that it is of a network type. This will be described using an example.

FIG. 2 is an explanatory diagram showing an example of a related thesaurus. In FIG. 2, nodes represent terms, and links between the nodes represent relationships. As shown in Fig. 2, in the related thesaurus, "official rate" and "interest rate" are related, and "interest rate" and "loan" are connected in a related relation. There is a related relationship. As described above, in the relation thesaurus, terms are associated with each other to form a network.

There are the following two conventional techniques for browsing such a related thesaurus. First, an "information retrieval device" described in Japanese Patent Application Laid-Open No. 5-233717
When the user inputs a term, a related term having a relevance within a range designated by the user is searched for and displayed. There are N levels of relevance depending on the level of relevance (eg, “large, medium, and small”). As a result, the user can efficiently refer to related terms having low relevance.

[0008] In the "data search device" described in Japanese Patent Application Laid-Open No. 9-44525, when a user inputs a term, the display position of the related term is determined according to the strength of the degree of association with the input term. More specifically, the present invention relates to a method of displaying a related thesaurus two-dimensionally with a network structure as shown in FIG. 2, in which terms having a higher degree of relevance to terms input by the user are displayed closer to the input terms. As a result, the user can visually grasp the degree of the degree of association with the input term.

[0009] These prior arts are useful when the user is interested in only a portion of the associated thesaurus. However, on the other hand, (1) it is not possible to provide the user with an overview of the related thesaurus, and (2) even if he is interested in a part of the related thesaurus, it takes time to reach a desired term. There is a problem. Less than,
Each will be described.

(1) The user cannot be provided with an overview of the related thesaurus: The above-described conventional technique displays only the related terms of the key terms input by the user. On the other hand, there is a user request to get an overview of the related thesaurus. For example, there are times when a user wants to refer to a related thesaurus, but does not know exactly what he / she wants to refer to without actually examining the related thesaurus.

In such a case, the user may first want to get an overview of the related thesaurus as to what terms are stored and then follow the partial structure of the network in the direction of interest. For example, in a hierarchical thesaurus, a user can get an overview by referring to a list of top-level terms. On the other hand, in the related art, the user needs to determine in advance which part of the network of the related thesaurus is to be referred to. for that reason,
It is not possible to respond to a user's request to browse the thesaurus after gaining an overview of the related thesaurus.

(2) It takes time to reach a desired term: As described above, the conventional technique requires input of a key term by a user. On the other hand, it is known that inputting an appropriate key is not easy. In particular, for a user who is not familiar with the related thesaurus to be searched, the term to be input is often a term that the user has come up with appropriately. For example, if you want to know about the economy,
"Economy" is entered as a key term, and so on.

[0013] When such a term is input as a key, it is unlikely that a related term matching the user's interest is directly obtained. Therefore, it will be necessary to select a key term from the displayed related terms, and further search and trace the related terms. As described above, since the related thesaurus is a network type thesaurus, the user may become "lost", that is, may not understand how the user has traced the thesaurus while following the related terms one after another. As described above, it is inefficient to check the thesaurus by appropriately adding a term that comes to mind.

[0014]

The problems to be solved are that the conventional technology cannot provide the user with the entire structure (outlook) of the related thesaurus and that the user is interested in a part of the related thesaurus. The point is that it takes time to reach the desired term even if you have it.

An object of the present invention is to provide a thesaurus browsing system and method capable of solving the problems of the prior art and improving the use efficiency of a related thesaurus, and a recording medium on which a processing program is recorded. is there.

[0016]

SUMMARY OF THE INVENTION To achieve the above object, the thesaurus browsing system and method of the present invention do not display only a part of a network type related thesaurus as in the prior art, but store the related thesaurus in the related thesaurus. A set of terms (representative terms) showing the overall structure (outline) of the related thesaurus is generated and displayed from the terms thus obtained. Then, when the user selects any of the sets, terms related to general terms belonging to each representative term of the selected set are extracted from the thesaurus, and some terms are similarly extracted from each of the extracted terms. Generate and display a set.

Thus, the user can be navigated from the general structure composed of general terms to the partial structure including more limited terms in the related thesaurus. For example, from the screen displaying a set of terms indicating an overview of the thesaurus as an initial state of navigation, the user selects a part of interest (a set of terms) and instructs zooming, so that the part of interest is displayed.
Reference can be made to detailed structures including more restrictive terms.

[0018]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a block diagram showing a first embodiment of the configuration of the thesaurus browsing system of the present invention according to the present invention. FIG. 2 shows the contents of a related thesaurus to be processed by the thesaurus browsing system in FIG. FIG. 3 is an explanatory diagram showing an outline of processing according to the present invention of the thesaurus browsing system in FIG. 1, and FIG. 4 is a block diagram showing an example of a hardware configuration of the thesaurus browsing system in FIG. It is.

First, the outline of the thesaurus browsing system of this embodiment will be described with reference to FIGS. The thesaurus browsing system of this example automatically generates a related thesaurus shown in FIG. 2 from a database of documents belonging to the economic field in particular, and gives an overview of a plurality of terms collected as general terms in FIG. (Term cluster) and displayed to the user.

Referring to FIG. 3, browsing of the related thesaurus in the economic field will be described. In FIG. 3, the top level is an overview of a thesaurus composed of a general set of terms. As shown in FIG.
It is composed of a set of terms including “tax system”, “income”, “National Tax Agency”, etc., and a set of terms including “BOJ”, “foreign currency”, “yen”, and “exchange”.

When the user selects a set of terms including “BOJ”, “foreign currency”, “yen”, and “exchange”, the terms related to the terms included in the selected set of terms are more limited. The set containing the appropriate terms is displayed. As an example, in the middle part of FIG. 3, a set of terms including “yen”, “yen sell”, “yen buy”, and “Tokyo forex market”, “foreign currency”, “dollar”, “IMF”, and “euro” are shown. A set of terms and the like are displayed.

If the user selects a set of terms that the user is more interested in, a set of terms composed of more limited terms can be referred to. That is,
In the lower middle of Fig. 3, "Yen buying", "Yen appreciation", "Trade surplus",
An example is a set of terms including "imbalance".

By referring to a set of terms composed of general terms in this way, the user can gain an overview of the related thesaurus. Furthermore, the thesaurus can be efficiently browsed by interactive processing (hereinafter, referred to as "zooming") for displaying a subset including more limited terms from a general term set selected by the user.

In order to realize such browsing of the thesaurus, in this embodiment, a thesaurus overview generating function and a term cluster zooming function are provided. The thesaurus overview generation function is for generating an overview of a related thesaurus, and includes the following two processing functions. That is, (a) a representative term extraction processing function for extracting a general term from the field corresponding to the relation stored in the thesaurus, and (b) a relation among the terms extracted by the representative term extraction processing function. This is a term cluster generation processing function for generating clusters by combining high-degree ones.

In the following, general terms in the field to which the thesaurus corresponds, terms that are elements of the thesaurus overview are “representative terms”, processing for generating a cluster from a set of terms is “term clustering”, and terms are constituent elements. The cluster that performs the operation is called a “term cluster”.

The zoom processing function of the term cluster is as follows.
A set including more limited terms is generated and displayed from a set of general terms, and includes the following two processing functions. That is, (a) when a user selects a term cluster, a related term acquisition processing function for acquiring a related term of a term belonging to the term cluster, and (b) a term cluster for clustering related terms acquired by the related term acquisition processing function. This is a generation processing function.

A hardware configuration for implementing each function for realizing such a thesaurus browsing will be described with reference to FIG. As shown in FIG. 4, the thesaurus browsing system according to the present embodiment includes a CPU (Central Pr
Ocessing Unit) 1, a hard disk 2, a memory 3, a display 4a, a display control unit 4b, a keyboard 5a, a keyboard control unit 5b, a mouse 6a, a mouse control unit 6b, and a bus 7.

The CPU 1 executes data the input / output, reading, storing, and various processes based on a program to perform a thesaurus browsing process according to the present invention. The hard disk 2 is a device for storing data, a memory 3
Is a device that loads and stores programs and data. The display 4a is a device that displays data to a user, and is controlled by the display control unit 4b. The keyboard 5a and the mouse 6a are devices for receiving input from a user,
b and the mouse controller 6b. Bus 7
Passes data between components.

In such a hardware configuration, a processing program relating to the thesaurus browsing method of the present invention is stored in a hard disk 2 from a recording medium such as an optical disk (not shown), loaded into a memory 3, and started up. The thesaurus browsing system shown in FIG.

The thesaurus browsing system in FIG. 1 is classified into the following three modules. That is, a processing unit that executes various types of processing (shown by rectangles in the figure),
A data storage unit (shown as a drum in the figure) for storing data, and data (shown as a parallelogram in the figure).
Hereinafter, each will be described.

First, as a processing unit, a related thesaurus generation unit 1a, a term vector extraction unit 1b, a representative term acquisition unit 1
c, a related term acquisition unit 1d, and a term cluster generation unit 1e. The related thesaurus generation unit 1a generates a related thesaurus from the document data stored in the document data storage unit 2a.

The term vector extraction section 1b extracts a term vector from the document data stored in the document data storage section 2a. The representative term acquisition unit 1c acquires a representative term from the term vectors stored in the term vector storage unit 2c. The related term acquisition unit 1d acquires a related term of a term belonging to the term cluster selected by the user. The term cluster generation unit 1e generates the representative term list 3
Cluster the terms stored in c or the related term list 3d. The detailed processing procedure of each of these units will be described later using a flowchart.

Next, there are four data storage units: a document data storage unit 2a, a related thesaurus storage unit 2b, a term vector storage unit 2c, and a thesaurus overview storage unit 2d. The document data storage 2a stores data of documents belonging to a certain field. For example, it is a newspaper article published in economic terms, or a specification of a patent given a certain classification number.

The related thesaurus storage unit 2b stores a related thesaurus generated from the document data stored in the document data storage unit 2a. The term vector storage unit 2c stores a term vector extracted from the document data stored in the document data storage unit 2a. Thesaurus overview storage 2
d stores an overview of the related thesaurus stored in the related thesaurus storage unit 2b.

Here, a detailed configuration of each data storage unit will be described with reference to FIGS. FIG. 5 is an explanatory diagram showing a configuration example of the document data storage unit in FIG. Figure 5
The document data storage unit 2a stores the document data 2a
01. The document data 2a01 stores text data of the document. The first line in FIG. 5 stores “What is the yen exchange rate of the Tokyo foreign exchange market on the 21st ...” as an example of a newspaper article on the economy.

FIG. 6 is an explanatory diagram showing a configuration example of the related thesaurus storage unit in FIG. As shown in FIG. 6, the relation thesaurus storage unit 2b includes a term X2b01, a term Y2b02, and a degree of relation 2b03. The terms X2b01 and Y2b02 store terms having a relational relationship, and the degree of relation 2b03 stores the degree of relation. In FIG. 6, the first line stores that "yen" and "Tokyo Forex Market" are related terms, and that the degree of relevance is "11.5". Note that the degree of association in this example is a value of the mutual information amount of a set of terms. This will be described later.

FIG. 7 is an explanatory diagram showing a configuration example of the term vector storage unit in FIG. As shown in FIG. 7, the term vector storage unit 2c includes a document ID 2c01 and an important term list 2c02. Document ID 2c
01 contains the I of the document stored in the document data storage unit 2a.
D is stored, and a list of important terms appearing in the document is stored in the important term list 2c02.
In FIG. 7, the first line stores that the important terms of the document with document ID 1 are “yen appreciation”, “Tokyo foreign exchange market”, “dollar depreciation” and the like.

FIG. 8 is an explanatory diagram showing an example of the configuration of the thesaurus overview storage unit in FIG. As shown in FIG. 8, the thesaurus overview storage unit 2d includes a term list 2d01. The term list 2d01 stores a list of terms belonging to the term cluster. In FIG. 8, the first line stores a term cluster including “economics”, “sales”, “consumer”, and the like. Note that the thesaurus overview shown in FIG. 8 is composed of ten term clusters as an example.

The data storage unit of the module configuration shown in FIG. 1 will be described next with reference to FIGS. As shown in FIG. 1, the data includes a co-occurrence term table 3a, a document term table 3b, a representative term list 3c, a related term list 3d, a correlation matrix 3e, an output term cluster list 3f, and an input term cluster list 3g.

The co-occurrence term table 3a stores a set of terms (a set of terms) that appear simultaneously within a certain range in the document data stored in the document data storage unit 2a. The document term table 3b is stored in the document data storage 2a.
The terms extracted from the documents stored in the document and their frequencies are stored for each document. The representative term list 3c stores the representative terms extracted from the term vector storage unit 2b. The related term list 3d stores the related terms of the terms stored in the input term cluster 3g.

The correlation matrix 3e stores the degree of association between the terms stored in the representative term list 3c or the related term list 3d. The output term cluster list 3f stores the result of clustering the terms stored in the representative term list 3c or the related term list 3d. The input term cluster list 3g stores the term cluster selected by the user.

Here, details of each data will be described with reference to FIGS. FIG. 9 is an explanatory diagram showing a configuration example of the co-occurrence term table in FIG. As shown in FIG. 9, the co-occurrence term table 3a stores the term X3a0
1, term Y3a02, and co-occurrence frequency 3a03. Term X3a01 and term Y3a02 include:
In the document data stored in the document data storage unit 2a, a set of terms appearing simultaneously within a certain range is stored. The co-occurrence frequency 3a03 stores the frequency at which the term set (set) appears at the same time.

Note that "co-occurrence" means that a term appears simultaneously with another term within a certain range, and "co-occurrence frequency" indicates the number of times that a term appears simultaneously. Further, a term that co-occurs with a certain term is called a “co-occurrence term”. Here, "same sentence" is used as an example of a certain range. As an example, in the first line of FIG. 9, “Yen market price” and “Tokyo forex market” are 2
It stores that it has appeared 59 times in the same sentence.

FIG. 10 is an explanatory diagram showing a configuration example of the document term table in FIG. As shown in FIG. 10, the document term table 3b includes a document ID 3b01, a term 3b02, and an appearance frequency 3b03.
The document ID 3b01 stores the ID of the document stored in the document data storage unit 2a. Term 3b02 contains the I
The term that appears in the document D and the appearance frequency 3b03 store the frequency of occurrence of the term in the document. In FIG.
The first line stores that “Tokyo Forex Market” appears twice in the document with the document ID “1”.

FIG. 11 is an explanatory diagram showing a configuration example of the representative term list in FIG. FIG. 11A shows a state in which important terms are acquired from the term vector, and FIG. 11B shows a list of states sorted after the acquisition of important terms. As shown in FIG. 11A, the representative term list 3c includes a representative term 3c01 and the number of documents 3c0.
2

The representative term 3c01 stores the representative term extracted from the term vector storage 2b. The number of documents 3c02 stores the number of documents in which the term has become an important term. In FIG. 11A, the first line stores that "exchange" was an important term in 451 documents.
FIG. 11 (b) shows a state in which the list of FIG. 11 (a) is sorted in descending order of the number of documents. In FIG. 11 (b), on the first line, "stock" is 498 documents and an important term. That is stored.

FIG. 12 is an explanatory diagram showing a configuration example of the related term list in FIG. FIG. 12A shows a state when the related terms of the terms belonging to the term cluster selected by the user are acquired, and FIG. 12B shows a list of states sorted after the acquisition of the related terms. FIG.
As shown in (a), the related term list 3d includes a related term 3d01 and a rank 3d02.

The related term 3d01 stores the related term of the term belonging to the term cluster selected by the user. In the order 3d02, an order indicating the order of the degree of association among the terms related to the term is stored.
In FIG. 12A, “Tokyo Forex Market” is stored as the first place in the fifteenth place. This indicates that the term “Tokyo Forex Market” has the highest relevance among the terms related to a term.

FIG. 12B shows a state in which the list shown in FIG. 12A is sorted in ascending order. FIG. 12B
In the first to tenth lines, the terms having the highest degree of relevance to the terms belonging to the input term cluster, that is, the terms of the first rank are stored.

FIG. 13 is an explanatory diagram showing a configuration example of the correlation matrix in FIG. As shown in FIG. 13, the correlation matrix 3
e is an n × n matrix. Here, n is the number of terms stored in the representative term list 3c or the related term list 3d. The element in the i-th row and the j-th column in the correlation matrix 3e shown in FIG. 13 is the degree of association between the i-th term and the j-th term in the representative term list 3c or the related term list 3d.

In FIG. 13, the degree of association “7.3” is stored in the second column of the first row. In the correlation matrix 3e,
The same value is stored in the diagonal elements of the matrix. That is, in FIG. 13, the same value “7.3” as in the first row and the second column is stored in the second row and the first column.

FIG. 14 is an explanatory diagram showing a configuration example of the output term cluster list in FIG. FIG. 14A shows an initial state of the output term cluster list 3f, and FIG.
FIG. 4B shows the state after the generation of the term cluster. FIG.
As shown in (a), the output term cluster list 3f is
Cluster ID 3f01 with maximum relevance, maximum relevance 3f0
2 and a term list 3f03.

Each line of the output term cluster list 3f is
It represents one term cluster, and the index of each row is the cluster ID. Cluster ID of maximum relevance
In 3f01, the ID of the term cluster having the highest degree of association with the term cluster is stored.
Stores the maximum relevance. The degree of association between term clusters will be described later. In the term list 3f03,
A list of terms belonging to the term cluster is stored.

As shown in FIG. 14A, the output term cluster list 3f stores a term cluster in which one term is stored in the term list 3f03 as an initial state. FIG. 14 shows a state in which the term clusters composed of these one terms are merged until a predetermined number is reached.
(B). The first line in FIG. 14B shows "I
A term cluster including “MF”, “dollar”, “foreign currency”, “euro”, and the like is stored.

FIG. 15 is an explanatory diagram showing a configuration example of the input term cluster list in FIG. As shown in FIG. 15, the input term cluster list 3g is the term list 3
g01. Output term cluster list 3f
Each line of the input term cluster list 3g represents one term cluster. The first line in FIG. 15 stores a term cluster including “yen”, “exchange”, “BOJ”, “foreign currency”, and the like.

The module configuration of the thesaurus browsing system of the present embodiment and the details of its components have been described above. Next, a detailed processing procedure of the thesaurus browsing system of this example will be described with reference to flowcharts shown in FIGS.

First, the overall processing procedure of the thesaurus browsing system will be described with reference to FIG. FIG.
FIG. 6 is a flowchart showing an example of a processing procedure of the thesaurus browsing method of the present invention.

As shown in FIG. 16, the thesaurus browsing method of the present embodiment is mainly composed of the following two processes. That is, a thesaurus browsing data generation process shown in FIG. 16A and a thesaurus browsing process shown in FIG. 16B. The former is a process for generating data for thesaurus browsing, and is executed in batches. On the other hand, the latter is a process for interactively responding to the user, and is executed in real time. Hereinafter, each of FIGS. 16A and 16B will be described, and then each step will be described using a detailed flowchart.

First, the thesaurus browsing data generation processing shown in FIG. FIG.
As shown in (a), in the thesaurus browsing data generation process, first, a related thesaurus is generated from document data (step 101), and term vectors of each document are extracted (step 102) to generate an overview of the related thesaurus. (Step 103).

Next, the thesaurus browsing process shown in FIG. 16B will be described. As shown in FIG. 16B, in the thesaurus browsing process, the user first displays the thesaurus overview stored in the thesaurus overview storage unit 2d (step 111), and then selects the displayed term cluster list by the user. If zooming is instructed (step 112), terms related to the terms belonging to the term cluster selected by the user are acquired (step 11).
3).

Then, they are clustered (step 114), and the generated term cluster is displayed to the user (step 115). If there is an instruction to end the thesaurus browsing from the user (step 116), the process is ended, and if not, the process returns to step 112.

Hereinafter, step 10 in FIG.
The first related thesaurus generation processing, the term vector extraction processing in step 102, thesaurus overview generation processing in step 103, and the related term acquisition processing in step 113 in FIG. 16B will be described in detail. Note that the term cluster generation processing in step 114 is the same as the term cluster generation processing included in the thesaurus overview generation processing shown in FIG.

First, a detailed procedure of the related thesaurus generation processing will be described. A process for extracting co-occurrence terms from document data and generating a related thesaurus is disclosed in Japanese Patent Laid-Open No. 5-282.
No. 367 (“Related keyword automatic generation device”) or Japanese Patent Application Laid-Open No. 8-161343 (“Related word dictionary creation device”). Here, as an example, a co-occurrence term from document data is described. The process of extracting and calculating the degree of association based on the mutual information amount will be described.

The mutual information amount is defined as "the information amount of another event Y estimated from the directly obtained event X (here, term)". Here, a set of terms (a set of terms) is used. Is considered as a normalized value. The equation for obtaining the mutual information is shown in Equation 1.

(Equation 1)

FIG. 17 is a flowchart showing a detailed processing procedure example of the related thesaurus generation processing in the thesaurus browsing data generation processing of FIG. 16 (a).
As shown in FIG. 17, in the related thesaurus generation processing, first, a variable i indicating the number of documents stored in the document data storage unit 2a is set to 1 and initialized (step 1a01), and i
While dividing the second document data into words, all the term sets appearing in the same sentence are extracted, and the co-occurrence term table 3a is extracted.
(Step 1a02).

At this time, if there is the same set of terms in the co-occurrence term table 3a, the co-occurrence frequency is increased by 1. If there is no such set of terms, the co-occurrence frequency is stored as 1. Then i is 1
The number is increased (step 1a03). If i is smaller than the number of documents (step 1a04), the process returns to step 1a02.

If i exceeds the number of documents, a variable j indicating the index of the co-occurrence term table 3a is initialized to 1 (step 1a05), and the degree of association of the j-th term set in the co-occurrence term table 3a is calculated. 1 and stores a set of terms having a degree of relevance equal to or greater than a certain threshold value in the relation thesaurus storage unit 2b (step 1a06). Next, j is incremented by 1 (step 1a07). If j is smaller than the number of term sets stored in the co-occurrence term table 3a (step 1a08), the process returns to step 1a06, and if j exceeds the co-occurrence number, processing is performed. To end.

Next, a detailed procedure of the term vector extraction processing will be described. FIG. 18 is a flowchart illustrating a detailed processing procedure example of the term vector extraction processing in the thesaurus browsing data generation processing of FIG. As shown in FIG. 18, the term vector extraction processing
First, a variable i indicating the number of documents stored in the document data storage unit 2a is set to 1 and initialized (step 1b0).
1) The i-th document is divided into words, and each word and the number of appearances are stored in the document term table 3b (step 1).
b02).

Then, i is increased by 1 (step 1b0
3) If i is smaller than the number of documents (step 1b04)
It returns to step 1b02. If i exceeds the number of documents, the number of appearing documents in the entire document database is counted for each of the terms stored in the document term table 3b (step 1b05), i is reset to 1 (step 1b06), and the i-th The term weights appearing in the document are calculated based on Equation 2, a term having a weight equal to or greater than a certain threshold is generated as an important term of the i-th document, and a term vector is stored in the term vector storage unit 2c (step 1b07). .

Further, i is increased by 1 (step 1b0
8) If i is smaller than the number of documents stored in the document data storage unit 2a (step 1b09), step 1b
Return to 07. If i exceeds the number of documents, the process ends.

Next, the detailed procedure of the thesaurus overview generation processing will be described. FIG. 19 is a flowchart illustrating a detailed processing procedure example of the thesaurus overview generation processing in the thesaurus browsing data generation processing of FIG.
As shown in FIG. 19, in the thesaurus overview generation process, first, a representative term which is an element of the thesaurus overview is extracted (step 1031), and the extracted representative terms are clustered to generate a term cluster (step 1032).
The generated term cluster is stored in the thesaurus overview storage unit 2d (step 1033).

Hereinafter, the representative term extraction processing in step 1031 and the term cluster generation processing in step 1032 will be described in detail. The procedure of the representative term extraction processing in the thesaurus overview generation processing will first be outlined, and then will be described in detail using a flowchart.

As clues for extracting general terms in a certain field, there are data such as the frequency of occurrence of terms in the relevant field and the number of documents appearing. Using these as they are, a method of extracting a term having a high appearance frequency or a term having a large number of appearing documents as a general term in the field can be considered, but a meaningless term may be extracted. For example, “economy” is expected to appear frequently in the economic field, but is unlikely to be meaningful.

On the other hand, in the representative term extraction processing of this embodiment, a meaningful general term is extracted by utilizing the importance of the term. Hereinafter, a specific description will be given. In the representative term extraction processing of this example, the “term vector” extracted in advance from a document database belonging to the relevant field
Use This "term vector" is a list of terms that characterize the document, and is described in "Salton, G., et al .:
A Vector Space Model for Automatic Indexing, Comm
unications of the ACM, Vol. 18, No. 11 (1975). "
t Frequency) can be extracted.

The tf · idf method is one of the most well-known document indexing methods, and includes the frequency of occurrence (tf) of a term in a certain document and the reciprocal (idf) of the number of documents in which the term appears. ) Is used as the importance of a term in the document, and a term having a high importance (hereinafter, important term) in the document is extracted and used as a term vector.

The equation for calculating the importance of a document is shown in equation (2).

(Equation 2) The denominator in the equation (2) is a value for normalizing the value of tf · idf by the size of the document in which the term appears.

In the representative term extraction processing, terms constituting a term vector are extracted, the number of documents each having an important term is counted, and terms that have become important terms in many documents are extracted as general terms. I do. next,
FIG. 20 shows a detailed processing procedure example of the representative term extraction processing.
This will be described with reference to FIG.

FIG. 20 is a flowchart showing an example of the detailed processing procedure of the representative term extraction processing. As shown in FIG. 20, in the representative term extraction processing, first, the document data storage unit 2
A variable i indicating the number of documents stored in a is set to 1 and initialized (step 1c01), important terms constituting the term vector of the i-th document are obtained from the term vector storage unit 2c, and important terms are acquired. If is already in the representative term list 3c, the number of documents 3c02 is incremented by one, and if not, the number of documents 3c02 is set to 1 and stored in the representative term list 3c (step 1c02).

Then, i is increased by 1 (step 1c0).
3) If i is smaller than the number of documents (step 1c04)
It returns to step 1c02. If i exceeds the number of documents, the representative term list 3c is sorted in descending order of the number of documents (step 1c05), leaving only the number specified from the top of the representative term list 3c, deleting important terms later, and processing. The process ends (step 1c06).

Next, as an explanation of the term cluster generation processing in the thesaurus overview generation processing, first, an outline of the term cluster generation processing will be described, and then a detailed description will be given using a flowchart. In order to generate a cluster from a set of terms, in the term cluster generation process, first,
The terms to be clustered are {t1, t2, ...,
tm}, and n clusters c1 = {t
1}, C2 = {t2},..., {Tm}.

Then, a pair of the term clusters having the highest degree of association is selected until the total number of clusters becomes equal to the designated number, and the process of merging into a single term cluster is repeated. Here, the degree of association rel (C,
D) includes, for example, a technique of setting the maximum value among the relevance R (t, s) of the term t belonging to C and the term S belonging to D to the relevance rel (C, D) of the term cluster. .

Next, a detailed processing procedure of the term cluster generation processing will be described with reference to FIG. FIG. 21 is a flowchart illustrating an example of a detailed processing procedure of the term cluster generation processing in the thesaurus browsing data generation processing of FIG.

As shown in FIG. 21, in the term cluster generation processing, first, a correlation matrix storing the degree of association between terms is generated (step 1e01). Next, a term cluster consisting of one term is generated, the output term cluster list 3f is initialized (step 1e02), and the number of terms to be clustered is set to a variable N indicating the number of term clusters to initialize (step 1e03).

A set of term clusters C1 and C2 having a high degree of association among the N term clusters are obtained (step 1e04), merged (step 1e05), and N is reduced by 1 (step 1e06). If N is not the same as the designated term cluster number (step 1e07), the process returns to step 1e04. If they are the same, the process ends.

Here, the correlation matrix generation processing in step 1e01, the term cluster initialization processing in step 1e02, the term cluster selection processing in 1e04, and the term cluster merge processing in step 1e05 will be described in detail with reference to flowcharts. . First, the correlation matrix creation processing in the term cluster generation processing will be described in detail.

FIG. 22 is a flowchart showing a detailed processing procedure example of the correlation matrix generation processing in the term cluster generation processing of FIG. In the following description of the flowchart of the correlation matrix generation process, the term list indicates a representative term list when the input is the representative term list 3c, and a related term list when the input is the related term list 3d.

As shown in FIG. 22, in the correlation matrix generation processing, first, a variable i indicating the row number of the correlation matrix is set to 1 and initialized (step 1e011), and the element of the ith row and the ith column of the correlation matrix is set. Is set to 0 (step 1e01)
2). Next, j indicating the column (or row) number of the correlation matrix is i +
1 is set (step 1e013), the degree of association between the i-th term and the j-th term in the term list is obtained from the association thesaurus (step 1e014), and the correlation matrix i
The value is set to the j-th row and the i-th row of the j-th row (step 1e015).

J is incremented by 1 (step 1e016),
If j exceeds the number of terms stored in the term list (step 1e017), the process proceeds to step 1e018, and if not, the process returns to step 1e014. Step 1e01
In step 8, i is incremented by 1. If i exceeds the number of terms stored in the term list (step 1e019), the process ends. If i does not exceed the number, the process returns to step 1e012.

Next, the term cluster initialization processing in the term cluster generation processing will be described in detail. FIG.
22 is a flowchart illustrating a detailed processing procedure example of a term cluster initialization process in the term cluster generation process of FIG. 21. As shown in FIG. 23, in the term cluster initialization processing, first, a variable i indicating the row number of the correlation matrix is set to 1 and initialized (step 1e021), and the ith term in the term list is output from the output term cluster list 3f. It is set in the i-th term list 3f03 (step 1e022).

The maximum value of the relevance stored in the i-th row of the correlation matrix is set to the i-th maximum value 3f02 of the output term cluster list 3f (step 1e023), and the number of the column having the maximum value is output. It is set to the i-th cluster ID 3f01 of the cluster list 3f (step 1e024). i is increased by 1 (step 1e02
5) If i exceeds the number of terms stored in the term list (step 1e026), the process ends; otherwise, the process returns to step 1e022.

FIG. 14 shows an example of a specific result of this processing. The example shown in FIG. 14 is based on the related term list 3d shown in FIG. That is, the term “IMF” in the first row in the related term list 3d is set in the term list 3f03 in the first row of the output term cluster list 3f, and the term “IMF” in the correlation matrix 3e is set to 1
The maximum relevance “14.4” in the row is set to the maximum relevance 3f02, and the column number “3” of the maximum relevance is set.
Is set to the cluster ID 3f03. Similarly, the second line of the output term cluster list 3f includes “dollar”,
“7.3” and “1” are replaced by “Euro” and “1” on the third line.
4.4 "and" 1 "are set.

Next, the term cluster selection processing in the term cluster generation processing will be described in detail. FIG.
22 is a flowchart illustrating a detailed processing procedure example of a term cluster selection process in the term cluster generation process of FIG. 21.

As shown in FIG. 24, in the term cluster selection processing, first, 1 is added to the variable i indicating the index of the output term cluster list 3f, and the variable C indicating the index of the term cluster to be selected as the most relevant set.
A variable M that stores 0 in each of 1 and C2 and the degree of association
ax is set to 0 to initialize (step 1e04)
1). Then, the maximum value stored in the i-th of the output term cluster list 3f is compared with the value of Max (step 1).
e042).

If Max is larger, step 1e04
Then, if Max is smaller, the i-th maximum value of the output term cluster list 3f is set to Max, i is set to C1, and the i-th cluster ID of the output term cluster list 3f is set to C2. (Step 1e04
3).

In step 1e044, i is increased by 1, and
If i exceeds the number of term clusters stored in the output term cluster list 3f (step 1e045), the process ends. If not, the process returns to step 1e042.

Next, the term cluster merge processing in the term cluster generation processing will be described in detail. FIG.
22 is a flowchart illustrating a detailed processing procedure example of a term cluster merge process in the term cluster generation process of FIG. 21. In the following description of the flowchart, C1 and C2 are the index of the term cluster to be merged.

As shown in FIG. 25, in the term cluster merge processing, first, a variable i indicating the column number of the correlation matrix 3e is set to 1
Is set and initialized (step 1e051). Next, the degree of association of the correlation matrix C1 row i column is compared with the degree of association of the correlation matrix C2 row i column (step 1e052). If the degree of association of the correlation matrix C2 row i column is smaller, the process proceeds to step 1e054; , Is set to the element of the correlation matrix C1 row i column (step 1e053).

For example, in the first row (C1) and the third row (C2) in the fourth column of the correlation matrix 3e in FIG. 13, the relevance is “4.1” and “8.6”, respectively. Correlation matrix 3
e, in the first row (C1) in the fourth column, "4.
"1" is deleted and set to "8.6", and the third line (C2)
Is set to “0”.

In step 1e054, 0 is set to the element in the 2nd row and the ith column of the correlation matrix C, and i is incremented by 1 (step 1e055). If i exceeds the number of terms in the term list (step 1e056), the process proceeds to step 1e057. If not, return to step 1e052.

In step 1e057, the terms stored in the output term cluster list C2 and the term stored in the second term list of the output term cluster list 3f are added to the C1 term list, and further stored in the C1 line of the correlation matrix 3e. The number of the column having the maximum value of the relevance is stored in the C1st “cluster ID 3f01 of the maximum relevance” in the output term cluster list 3f, and the value is stored in the “maximum relevance 3f02”. Then, the information of the C2th “term list 3f03” of the output term cluster list 3f is deleted, and the “term cluster ID 3f01 having the maximum value” is deleted.
Is set to "0" for "maximum relevance 3f02" (step 1e058), and the process ends.

For example, in the output term cluster list 3f of FIG. 14, the first (first line) term list 3f
03 (“IMF”) “$” in the term list on the second line
Is added, and “3” is added to the cluster ID 3f01 of the maximum relevance.
, And “14.4” in the maximum relevance 3f02,
“Dollar” in the term list on the second line is deleted, and “0” is stored in the cluster ID 3f01 and the maximum relevance 3f02 of the maximum relevance. This operation is repeated, and the term list is deleted and the cluster ID 3f01 having the highest relevance is set.
By deleting the row in which “0” is stored in the maximum relevance 3f02, the content shown in FIG. 14B is obtained.
Note that the number of terms in each row is a predetermined number such as "10".

Next, an outline of the related term acquisition processing in the thesaurus browsing processing will be described first, and then details will be described with reference to a flowchart. The related terms of the terms belonging to the term cluster selected by the user from the input term cluster list 3g (hereinafter, referred to as “seed terms”) can be obtained by searching the related thesaurus.

If the total number of acquired related terms is small (for example, within 300), all of them may be subjected to clustering. On the other hand, when the number is large, related terms are equally acquired from various terms as an example. That is, the related terms are acquired from all the seed terms in the order of the degree of relevance, X terms at a time. As a result, the number of terms to be clustered falls within a range of a predetermined number, so that it is possible to suppress the amount of calculation for generating the term cluster.

The detailed processing procedure of the related term acquisition processing will be described with reference to FIG. FIG. 26 shows the state shown in FIG.
It is a flowchart which shows the detailed example of a processing procedure of the related term acquisition process in the solar browsing process of FIG. As shown in FIG. 26, in the related term acquisition processing, first, the terms stored in the term list 3g01 of the input term cluster list 3g are stored in the related term list 3d.
Here, the rank 3d02 of the row storing each term is set to 1 (step 1d01). For example, "Tokyo Forex Market" on line 15 and "Dollar" on line 20 are shown.

The variable i representing the total number of terms stored in the term list 3g01 of the input term cluster list 3g is set to 1 for initialization (step 1d0).
2), term list 3 of input term cluster list 3g
Among the terms stored in g01, the related terms of the i-th term are obtained by searching the related thesaurus storage unit 2b, and they are sorted and ranked in descending order of the degree of relevance, and placed at the end of the related term list 3d. Store (step 1d0
3). As a result, for example, lines 15 to 18 in the related term list 3d in FIG.

Further, i is increased by 1 (step 1d0
4). If i exceeds the total number of terms stored in the term list 3g01 (step 1d05), step 1d0
Go to step 6, if not exceeded, return to step 1d03. In step 1d06, the related term list 3d is sorted in ascending order of rank. As a result, the related term list 3d shown in FIG. 12B is obtained.
The following processing is performed.

That is, 1 is set to a variable i indicating the index of the related term list 3d, and 0 is set to a variable j indicating the number of terms selected as a clustering target among the related terms.
Is set and initialized (step 1d07). If the i-th term in the related term list 3d is not the first to i-1st terms (step 1d08), j is incremented by 1 (step 1d09), and the process proceeds to step 1d11. If it is in the related term list 3d, the i-th order of the related term list 3d is set to 0 (step 1d1).
0). In step 1d11, i is incremented by 1 (step 1d11). If i exceeds the number of related terms stored in the related term list 3d (step 1d12), the process proceeds to step 1d15.

[0108] Otherwise, i- in the related term list 3d
If the first order and the i-th order are the same (step 1d13), the process returns to step 1d08. If they are not the same, the process proceeds to step 1d14, and if j becomes equal to the number of terms to be clustered, the i-th and related terms and the related terms whose rank is 0 are deleted (step 1d15),
The process ends. However, if j is not the same as the number of terms to be clustered, the process returns to step 1d08. In this way, the related term list 3d in FIG. 12B has no duplication.

The processing procedure of the thesaurus browsing method of this example has been described above. Hereinafter, a specific screen related to such thesaurus browsing will be described. FIG. 27 is an explanatory diagram illustrating a configuration example of a screen displayed by the thesaurus browsing system in FIG. 1.
This screen is an example in which the related thesaurus in the economic field shown in FIG. 6 is being browsed. FIG. 27A shows a display state of the thesaurus overview, and FIG. 27B shows a zooming state of the term cluster.

As shown in FIG. 27A, an overview of the thesaurus in the economic field includes term clusters composed of “economics”, “sales”, “consumers”, etc., “tax system”, “income”, “National Tax Agency”, etc. Are displayed. Here, when the user is interested in the third term cluster, selects the term cluster, and clicks the <zoom> button to instruct zooming, the screen changes to the state shown in FIG. 27B.

FIG. 27B shows a state in which the third term cluster in FIG. 27A is selected and zoomed.
A term cluster composed of “IMF”, “dollar”, “foreign currency”, etc., a term cluster composed of “yen”, “Tokyo forex market”, “yen purchase”, etc. are displayed. In this manner, the user can browse the related thesaurus efficiently without inputting a specific term.

Next, first to third modified examples of such an embodiment will be described. First, a modified example of the representative term extraction processing will be described as a first modified example. The hardware configuration of the thesaurus browsing system according to the first modification is the same as that of the embodiment shown in FIG. The module configuration is the same as the module configuration in FIG. 1 except that the term vector extraction unit 1b and the term vector storage unit 2b are omitted.

In the first modification, the representative term extraction unit 1
In a modified example of c, a representative term is extracted by using constituent elements of a document as clues. Specifically, terms are extracted from constituent elements in which terms that characterize documents tend to appear, and terms having a large number of appearing documents are stored as representative terms in the representative term list 3c. For example, if the data stored in the document data storage unit 2a is a newspaper article, the term appearing in the first sentence, that is, the headline, is extracted. Terms are extracted from the terms to be important terms.

According to the modified example of the representative term extracting unit 1c in the first modified example, in the case of a newspaper article in which the components of a document including general terms are clear, a thesaurus overview is efficiently generated. be able to.

Next, as a second modified example, a thesaurus browsing system including a process of editing a term cluster by a user will be described. The hardware configuration and module configuration of the thesaurus browsing system according to the second modified example are the same as those shown in FIGS. In the second modified example, the input term cluster 3g input to the related term acquisition unit 1d is one in which the user selects and adds / deletes a term from the displayed term clusters.

For example, in the screen example shown in FIG.
An "edit" button is provided next to the "zoom" button, and when the user selects the third term cluster ("yen, foreign exchange, BOJ, foreign currency", ...) and the "edit" button, the third term cluster is displayed.
Terms in the second term cluster ("Yen", "Exchange",
"BOJ", "foreign currency", ...)) and "Add"
An "edit screen" provided with a button and a "delete" button is displayed in another window.

The user clicks “Add” on this “edit screen”.
Operate the button and the "Delete" button to select only the necessary terms from the list, such as "Yen", "Exchange",
Create a new term cluster leaving only the "BOJ". Then, when the user selects the “end” button or the like,
Perform zooming on the new term cluster.

As described above, according to the second modification, the user can edit the input term cluster 3g, the user can zoom the thesaurus in accordance with his / her interest, and the user-friendly thesaurus can be used. Browsing can be achieved.

Further, as a third modification, when displaying the zooming result of the term cluster selected by the user,
A thesaurus browsing technique for displaying seed terms (terms belonging to the selected original term cluster) in a state that is distinguishable from other related terms is described.

The outline of the hardware configuration and processing procedure of the thesaurus browsing system according to the third modification is the same as that described with reference to FIGS. Book 3
Of the output term cluster 3f
Is displayed in a different color or font so that the seed term can be distinguished from other terms. Here, the seed term is a term belonging to the term cluster that has been input for term clustering, and belongs to any of the term clusters selected by the user from the thesaurus overview to the currently displayed term cluster.

In this way, unlike a hierarchical thesaurus in which browsing histories can be displayed hierarchically,
Even in a network-type thesaurus that is easily lost, the user can easily know what terms have been traced so far only by referring to the current state.

The thesaurus browsing system and method for browsing a related thesaurus automatically generated from document data belonging to a certain field have been described.
Next, as a second embodiment of the present invention, a thesaurus browsing technique when a user inputs a key term will be described.

First, the module configuration of the thesaurus browsing system according to the second embodiment will be described with reference to FIG. 28. Next, the processing procedure of the thesaurus browsing method will be described with reference to FIG. 29, and finally, with reference to FIG. 13 shows an example of a thesaurus browsing screen according to the second embodiment. The hardware configuration of the thesaurus browsing system according to the second embodiment may be the same as that of the first embodiment shown in FIG.

FIG. 28 is a block diagram showing a second embodiment of the thesaurus browsing system according to the present invention. As shown in FIG. 28, the present thesaurus browsing system includes a related term acquisition unit 1d and a term cluster generation unit 1e as processing units, a related thesaurus storage unit 2b as a data storage unit, a term 3h as data, a related term list 3d, and a correlation matrix. 3e and an output term cluster list 3f.

Among these, the term 3h is a term input by the user. The outline of the other components is the same as the example shown in FIG. The processing operation of the thesaurus browsing system having such a configuration will be described with reference to FIGS.

FIG. 29 is a flowchart showing an example of a processing procedure according to the present invention of the thesaurus browsing system in FIG. As shown in FIG. 29, first, the related thesaurus is searched, the related term of the term 3h input by the user is acquired and stored in the related term list (step 201), and the terms stored in the related term list are clustered. (Step 202), the term cluster is displayed to the user (Step 203).

Of these processes, the related term acquisition process in step 201 is the same as the related term acquisition process in the first embodiment if the user selects a term cluster consisting of one term. The description is omitted. Similarly, the term cluster generation processing in step 202 is the same as the term cluster generation processing in the first embodiment, and a description thereof will be omitted.

FIG. 30 is an explanatory diagram showing a configuration example of a screen displayed by the thesaurus browsing system in FIG. The screen shown in FIG. 30 is an example in which the key term “yen appreciation” is input to the related thesaurus in the economic field shown in FIG. 6 for browsing. As shown in FIG. 30, the terms related to “yen appreciation” include “dollar depreciation”, “Tokyo foreign exchange market”, and the like.

For example, as the first term cluster, “dollar depreciation”, “trade imbalance”, “reduction of surplus”, etc., as the second term cluster, “Tokyo foreign exchange market”, “yen buying”,
You can see that there is a “margin”. As described above, the related terms are divided into several term clusters and displayed, so that the user can efficiently refer to the related terms. As described above, the browsing of the related thesaurus when the user inputs a term has been described as the second embodiment.

As described above with reference to FIGS. 1 to 30, the thesaurus browsing system and method according to the present embodiment displays an overview (overall structure) of a thesaurus composed of general terms. Navigate the user from the overview to a substructure that contains more specific terms. Thus, the user can easily grasp the outline of the related thesaurus, and can browse the related thesaurus efficiently by zooming the term cluster in which the user is interested.

That is, even if a user cannot input an appropriate search term because the related thesaurus search request is vague or the user is not familiar with the related thesaurus to be searched, the user can refer to the displayed overview of the thesaurus. By grasping the overall structure and selecting a more interesting term cluster and instructing zooming, the thesaurus can be browsed. As a result, it is difficult to grasp the entire structure, and it is possible to efficiently refer to a network-type related thesaurus that is easily lost during browsing.

When the related thesaurus is generated from the document database, it is possible to browse the database itself, which cannot be performed by the conventional document database search. That is, conventionally, as a technique for referring to a document database, there are a technique of searching for a stored document to refer to a text, and a technique of referring to a list of stored bibliographic information such as titles. The user refers to data corresponding to each of the stored documents, and the user wants to grasp the overall characteristics rather than the individual data stored in the document database (for example, the user wants to know the tendency of the patent database in a certain year). It was not a useful technique for).

On the other hand, as in the present example, a related thesaurus automatically generated from a document database can extract a characteristic related relationship in the database, so that browsing the related thesaurus enables the related thesaurus to be extracted. Features can be grasped. For example, in a related thesaurus generated from a patent database for a certain year, it is possible to grasp that the degree of relevance between “voice recognition” and “car navigation” is strong. Further, in the related art, only the partial structure of the related thesaurus is displayed, but in this example, the entire related thesaurus can be browsed.

The present invention is not limited to the embodiment described with reference to FIGS. 1 to 30 and can be variously modified without departing from the gist thereof. For example, in this example, a related thesaurus of document data in the economic field has been described, but the present invention can be similarly applied to a related thesaurus such as a technical document database or a patent database.

[0135]

According to the present invention, it is possible to provide a user with an overview of a thesaurus composed of general terms in a related thesaurus, and to display a partial structure of interest to the user in more detail. It is possible to provide the user with a new information processing technology called “browsing of a document database” different from the conventional document database search, and the user can grasp the overall characteristics of the document database from the displayed thesaurus, Further, by zooming in the part of interest, it is possible to know the detailed characteristics of the document database, and it is possible to improve the use efficiency of the related thesaurus.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a first embodiment of a configuration according to the present invention of a thesaurus browsing system of the present invention.

FIG. 2 is an explanatory diagram illustrating an example of a related thesaurus.

FIG. 3 is an explanatory diagram showing an outline of a process according to the present invention of the thesaurus browsing system in FIG. 1;

FIG. 4 is a block diagram illustrating a hardware configuration example of the thesaurus browsing system in FIG. 1;

FIG. 5 is an explanatory diagram showing a configuration example of a document data storage unit in FIG. 1;

FIG. 6 is an explanatory diagram showing a configuration example of a related thesaurus storage unit in FIG. 1;

FIG. 7 is an explanatory diagram showing a configuration example of a term vector storage unit in FIG. 1;

8 is an explanatory diagram illustrating a configuration example of a thesaurus overview storage unit in FIG. 1. FIG.

FIG. 9 is an explanatory diagram showing a configuration example of a co-occurrence term table in FIG. 1;

FIG. 10 is an explanatory diagram showing a configuration example of a document term table in FIG. 1;

FIG. 11 is an explanatory diagram showing a configuration example of a representative term list in FIG. 1;

FIG. 12 is an explanatory diagram showing a configuration example of a related term list in FIG. 1;

FIG. 13 is an explanatory diagram showing a configuration example of a correlation matrix in FIG. 1;

FIG. 14 is an explanatory diagram showing a configuration example of an output term cluster list in FIG. 1;

FIG. 15 is an explanatory diagram showing a configuration example of an input term cluster list in FIG. 1;

FIG. 16 is a flowchart illustrating a processing procedure example of the thesaurus browsing method of the present invention.

FIG. 17 is a flowchart illustrating a detailed processing procedure example of a related thesaurus generation process in the thesaurus browsing data generation process of FIG.

FIG. 18 is a flowchart illustrating a detailed processing procedure example of a term vector extraction process in the thesaurus browsing data generation process of FIG. 16A.

FIG. 19 is a flowchart illustrating a detailed example of a thesaurus overview generation process in the thesaurus browsing data generation process of FIG. 16A.

FIG. 20 is a flowchart illustrating a detailed processing example of a representative term extraction process.

FIG. 21 is a flowchart illustrating a detailed processing procedure example of a term cluster generation process in the thesaurus browsing data generation process of FIG. 16A.

FIG. 22 is a flowchart illustrating a detailed processing example of a correlation matrix generation process in the term cluster generation process of FIG. 21;

23 is a flowchart illustrating a detailed processing procedure example of a term cluster initialization process in the term cluster generation process of FIG. 21;

FIG. 24 is a flowchart illustrating a detailed processing procedure example of a term cluster selection process in the term cluster generation process of FIG. 21;

FIG. 25 is a flowchart illustrating a detailed processing procedure example of a term cluster merge process in the term cluster generation process of FIG. 21;

FIG. 26 is a flowchart showing a detailed processing procedure example of a related term acquisition process in the solar browsing process of FIG. 16 (b).

FIG. 27 is an explanatory diagram showing a configuration example of a screen displayed by the thesaurus browsing system in FIG. 1;

FIG. 28 is a block diagram showing a second embodiment of the configuration of the thesaurus browsing system of the present invention according to the present invention.

FIG. 29 is a flowchart illustrating an example of a processing procedure according to the present invention of the thesaurus browsing system in FIG. 28;

FIG. 30 is an explanatory diagram showing a configuration example of a screen displayed by the thesaurus browsing system in FIG. 28;

[Explanation of symbols]

1: CPU, 2: Hard disk, 3: Memory, 4a:
Display, 4b: display controller, 5a: keyboard, 5b: keyboard controller, 6a: mouse, 6
b: mouse control unit, 7: bus, 1a: related thesaurus generation unit, 1b: term vector extraction unit, 1c: representative term acquisition unit, 1d: related term acquisition unit, 1e: term cluster generation unit, 2a: document data storage Section, 2b: related thesaurus storage section, 2c: term vector storage section, 2d: thesaurus overview storage section, 3a: co-occurrence term table, 3b:
Document term table, 3c: representative term list, 3d:
Related term list, 3e: correlation matrix, 3f: output term cluster list, 3g: input term cluster list, 3
h: term, 2a01: document data, 2b01: term X, 2b02: term Y, 2b03: degree of association, 2c0
1: Document ID, 2c02: Important term list, 2d0
1: term list, 3a01: term X, 3a02: term Y, 3a03: co-occurrence frequency, 3b01: document ID, 3
b02: Term, 3b03: Frequency of appearance, 3c01: Representative term, 3c02: Number of documents, 3d01: Related term, 3
d02: ranking, 3f01: cluster ID of maximum relevance,
3f02: maximum relevance, 3f03: term list, 3g
01: Term list.

Continued on the front page (72) Inventor Yasushi Morimoto 1-280 Higashi Koigakubo, Kokubunji-shi, Tokyo Inside the Hitachi, Ltd. Central Research Laboratory (72) Inventor Noriyuki Yamazaki 5030 Totsuka-cho, Totsuka-ku, Yokohama-shi, Kanagawa Prefecture Software of Hitachi, Ltd. Within the Business Unit (72) Inventor Keiko Iida 5030 Totsuka-cho, Totsuka-ku, Yokohama-shi, Kanagawa Prefecture Within the Software Division, Hitachi, Ltd. (72) Inventor Yasuhiko Uchida 5030 Totsuka-cho, Totsuka-ku, Yokohama-shi, Kanagawa Software, Ltd. F-term in business division (reference) 5B075 ND02 ND35 NR02 NR12 PP13 PQ02 PQ42 PQ46 QP03 5E501 AC33 AC34 BA05 CA02 CB09 EA10 EB05 EB20 FA03 FA06

Claims (15)

[Claims] 1. A thesaurus browsing system for searching and displaying a thesaurus storing a plurality of terms having a relational relationship, wherein an overview term cluster obtained by extracting and clustering a plurality of general terms representing an overview of the thesaurus is provided. A thesaurus overview generating means for displaying, and a thesaurus zooming means for displaying a limited term cluster obtained by clustering by extracting terms related to each of the general terms belonging to the overview term cluster selected by the user. Thesaurus browsing system. 2. The thesaurus browsing system according to claim 1, wherein terms related to each of the limited terms belonging to the limited term cluster selected by the user are extracted and clustered, and a newly obtained limited term cluster is obtained. A thesaurus browsing system comprising a second thesaurus zooming means for displaying. 3. The thesaurus browsing system according to claim 1 or 2,
The thesaurus overview generating means is a representative term obtaining means for extracting the representative terms that are the general terms from the terms having a relational relationship, and collectively collects those having a high degree of association from the representative terms extracted by the representative term obtaining means. A term cluster generating means for generating an overview term cluster. 4. The thesaurus browsing system according to claim 1, wherein said thesaurus zooming means extracts a term related to a general term belonging to an overview term cluster selected by said user. Related term acquisition means for extracting terms from the thesaurus, term cluster generation means for acquiring the degree of relevance between terms extracted by the related term acquisition means from the thesaurus, and summing terms having a high degree of relevance to generate the limited term cluster; And a thesaurus browsing system comprising: 5. A thesaurus browsing system for searching and displaying a thesaurus storing a plurality of terms having a related relationship, wherein related terms acquiring means for extracting terms related to a term input by a user from the thesaurus. And a term cluster generating means for obtaining a degree of relevance between terms extracted by the related term obtaining means from the thesaurus and generating a term cluster by summing up terms having a high degree of relevance, wherein the term cluster generating means A thesaurus browsing system characterized by displaying the generated term cluster. 6. The thesaurus browsing system according to claim 5, further comprising: a thesaurus zooming means for extracting terms related to each of the terms belonging to the term cluster selected by the user and displaying a limited-term cluster obtained by clustering. Thesaurus browsing system characterized by the following. 7. A thesaurus browsing method for a device for searching and displaying a thesaurus storing a plurality of terms having a related relationship, wherein a plurality of general terms representing an overview of the thesaurus are extracted and clustered. A thesaurus overview generation processing step of displaying a cluster; and a thesaurus zooming processing step of extracting terms related to each general term belonging to the overview term cluster selected by the user and displaying a limited-term cluster obtained by clustering. A thesaurus browsing method characterized by the above-mentioned. 8. The thesaurus browsing method according to claim 7, wherein terms related to each of the limited terms belonging to the limited term cluster selected by the user are extracted and clustered, and a newly obtained limited term cluster is obtained. A thesaurus browsing method comprising a second thesaurus zooming processing step for displaying. 9. The thesaurus browsing method according to claim 7, wherein the thesaurus overview generating step extracts a representative term that is the general term from terms having a related relationship. A thesaurus browsing method, comprising: a representative term acquisition processing step; and a term cluster generation processing step of generating the overview term cluster by combining those having high relevance from the representative terms extracted in the representative term acquisition processing step. . 10. The thesaurus browsing method according to any one of claims 7 to 9, wherein the thesaurus zooming processing step includes the step of recognizing a term that is related to a general term belonging to an overview term cluster selected by the user. Related term acquisition processing step extracted from the thesaurus, and the degree of association between the terms extracted in the related term acquisition processing step is acquired from the thesaurus,
And a term cluster generation processing step of generating the limited term cluster by combining terms having a high degree of relevance. 11. The term browsing method according to any one of claims 7 to 10, wherein a term cluster editing processing step of adding / deleting a term belonging to the overview term cluster based on an editing instruction from a user. And performing a process in the thesaurus zooming process step on the overview term cluster selected and edited by the user. 12. The thesaurus browsing method according to claim 7, wherein the general term cluster selected by the user is selected from the terms belonging to the limited term cluster displayed in the thesaurus zooming processing step. A thesaurus browsing method, comprising a term cluster display processing step of identifying and displaying the terms to which the user belongs. 13. A thesaurus browsing method for a device for searching and displaying a thesaurus storing a plurality of terms having a related relationship, wherein the term related to the term input by a user is extracted from the thesaurus. Acquisition processing step, acquiring the degree of relevance between terms extracted by the related term acquisition means from the thesaurus,
A term cluster generating step of generating a term cluster by grouping terms having a high degree of relevance, and displaying the term cluster generated by the term cluster generating means. 14. The thesaurus browsing method according to claim 13, further comprising a thesaurus zooming processing step of extracting terms related to each of the terms belonging to the term cluster selected by the user and displaying a limited-term cluster obtained by clustering. A thesaurus browsing method characterized by the above-mentioned. 15. A recording medium for recording a computer-readable program and data, wherein the program records a program for causing a computer to execute the processing of each step according to any one of claims 7 to 14. A recording medium characterized by the above-mentioned.