JP2000285122A - Apparatus and method for generating a thesaurus and recording medium storing a thesaurus generating program - Google Patents

Abstract

(57) [Summary] [Purpose] The burden of registering words in a thesaurus and registering data indicating semantic relationships is reduced. A sentence input using an input device is analyzed by morphological analysis to extract words (nouns). If the extracted word is not yet registered in the thesaurus 10, the word is newly registered in the thesaurus 10. The inclusion relation of the character codes is determined between words newly registered in the thesaurus 10 and words already registered in the thesaurus 10. Depending on the determination result, the superordinate (BT) or subordinate concept (N) between the newly registered word and the registered word
Relationship data indicating the relationship T) is registered in the thesaurus 10.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a thesaurus generation apparatus and method, and a recording medium on which a thesaurus generation program is recorded.

[0002]

[Background] In a thesaurus, many words (terms, keywords) are stored, and semantic relationships between words and other words, for example, words of a superordinate concept, words of a subordinate concept, synonymous words, and related words. Data indicating a relationship such as a word is stored.

[0003] Such a thesaurus can be used in search processing of documents and the like stored in a database. By using a thesaurus, not only can documents be searched based on key words (index terms),
You can also search for documents that contain higher or lower concept words, synonyms, and related words of the key word.

[0004] When creating a thesaurus, a developer or a user registers a large number of words using an input device, and further inputs data representing the relation between a broader concept word, a lower concept word, synonyms and related words into the input device. Register using Therefore, the operation of registering a large number of words in the thesaurus and the operation of inputting data representing the semantic relationship of the registered words require considerable time and effort.

[0005]

DISCLOSURE OF THE INVENTION The present invention compares the operation of registering a word in the thesaurus with the operation of registering data indicating the relationship between words newly registered in the thesaurus and words already registered in the thesaurus. It is an object of the present invention to provide an apparatus and method for generating a thesaurus which can simplify the target.

[0006] A thesaurus generation device according to the present invention is a storage means for storing, for a registered word and at least a part of the registered word, a thesaurus describing the relationship between the word and another word, and a sentence. Input means for inputting a word, word extracting means for extracting a word included in the sentence input by the input means, and determining whether the word extracted by the word extracting means is a word registered in a thesaurus A registration judging means, a registration means for newly registering the extracted word in the thesaurus when the extracted word is judged to be a word not registered in the thesaurus, a character of the newly registered word newly registered in the thesaurus Compares a column with a character string of a registered word already registered in the thesaurus, and determines the inclusion relation between them If it is determined by the engagement determining means and the inclusion relation determining means that the character string of the newly registered word includes the character string of the registered word, the registered word is a superordinate term of the newly registered word. The relationship that there is and the relationship that the newly registered word is a lower concept word of the above registered word is registered in the thesaurus, and when it is determined that the character string of the newly registered word is included in the character string of the already registered word, Is provided with a relationship registering means for registering, in a thesaurus, a relationship that the registered word is a new concept word and a relationship that the newly registered word is a higher concept word of the registered word. .

[0007] A thesaurus generation method according to the present invention is a method for generating a thesaurus describing a relationship between a registered word and at least a part of the registered word, and the relationship between the word and another word. , Extract words included in a given sentence, and determine whether the extracted words are words registered in a thesaurus,
If it is determined that the extracted word is not a word registered in the thesaurus, the extracted word is newly registered in the thesaurus, and the character string of the newly registered word newly registered in the thesaurus and already registered in the thesaurus. The registered words are compared with the registered words, and their inclusion relations are determined. If it is determined that the newly registered words include the registered words, Is registered in the thesaurus that the newly registered word is a superordinate concept word of the newly registered word and the relationship that the newly registered word is a lower concept word of the above registered word, and the character string of the newly registered word is included in the character string of the already registered word. If it is determined that the registered word is a new concept word is a lower concept word, and the relationship that the newly registered word is a higher concept word of the registered word is thesaurus. It is intended to be registered in.

The newly registered word is extracted from a Japanese sentence by, for example, morphological analysis, and the inclusion relation between the newly registered word and the registered word is determined using those character strings.

According to the present invention, when a word extracted from a sentence is not yet registered in the thesaurus, the extracted word is newly registered in the thesaurus. The inclusion relation between the character string of the newly registered word and the character string of the registered word is determined. If the registered word includes a higher concept word or a lower concept word of the newly registered word, the newly registered word and such (higher level) The relation between the concept word or the subordinate concept word and the registered word is registered in the thesaurus.

An example will be given for clarity. It is assumed that "shoes" have already been registered in the thesaurus and that "exercise shoes" have been newly registered. The character string representing the newly registered word “exercise shoes” includes the character of the already registered word “shoes”. "Shoes" means "athletic shoes", "women's shoes", "men's shoes",
It is positioned as a superordinate term for various words such as "climbing shoes" and "shoehorns". As described above, when it is determined that the character string of the newly registered word (shoes) includes the character string of the registered word (shoes), the registered word (shoes) is added to the newly registered word (shoes). ) And the relationship that the newly registered word (exercise shoes) is a lower concept word of the registered word (shoes) is registered in the thesaurus.

Conversely, it is assumed that "exercise shoes" have already been registered in the thesaurus and that "shoes" have been newly registered. The character string representing the newly registered word “shoes” is included in the character that is already registered word “exercise shoes”. As described above, when it is determined that the character string of the newly registered word (shoes) is included in the character string of the registered word (exercise shoes), the registered word (shoes) is determined.
The relationship that is a lower concept word of the newly registered word (shoes) and the relationship that the newly registered word (shoes) is a higher concept word of the already registered word (exercise shoes) are registered in the thesaurus.

As described above, according to the present invention, new words are sequentially registered in the thesaurus in accordance with the input text, and the new words are registered between the newly registered words and the words already registered in the thesaurus. In, the relationship between the upper concept word and the lower concept word is registered. The burden of registering a word in the thesaurus and the burden of registering data indicating the relationship between a higher concept word or a lower concept word are reduced.

The present invention further provides a recording medium recording a thesaurus generation program for generating a thesaurus.

A recording medium storing a thesaurus generation program according to the present invention creates a thesaurus describing a relationship between a registered word and at least a part of the registered word, the relationship between the word and another word. A program that determines whether a given word is a word registered in a thesaurus and, if it is determined that the word is a word not registered in the thesaurus, converts the word into a thesaurus The newly registered word is compared by comparing the character string of the newly registered word newly registered in the thesaurus with the character string of the already registered word already registered in the thesaurus. Is determined to include the character string of a registered word, the relationship that the registered word is a superordinate concept word of the newly registered word. And the relationship that the newly registered word is a lower concept word of the registered word is registered in the thesaurus, and if it is determined that the character string of the newly registered word is included in the character string of the registered word, The computer is controlled so that the relation that the registered word is a lower concept word and the relation that the newly registered word is a higher concept word of the registered word is registered in a thesaurus.

According to this program, for a given word using a computer, registration of the word,
Data indicating a higher-order or lower-order relationship between the word and the registered word is automatically registered.

The recording medium includes an optical disk, a magneto-optical disk, a magnetic disk, a magnetic tape, a semiconductor memory, and the like.

The present invention can be extended more generally to determination and registration of the inclusion relation of a data structure. According to the present invention, there is provided an apparatus for storing code data and an inclusion relation between the code data, a means for inputting new code data, and registering the new code data input by the input means in the storage means. Code data registering means, means for determining the inclusion relationship between newly input code data and code data already registered in the storage means, and the inclusion relationship of the code data determined to have an inclusion relationship. It is provided with inclusion relation registering means for registering in the storage means in association with the code data.

Code data includes character (string) code data representing a word, code data representing a number and other symbols,
Code data representing chemical formulas, code data representing various materials, products, parts, etc., code data representing structural relationships, other relationships, and the like are included.

According to the present invention, the inclusion relation between the inputted code data or between the inputted code data and the already registered code data is automatically registered. That is, if it is determined that there is an inclusion relationship, the code data and data representing the inclusion relationship with the already registered code data are registered.

The present invention also provides a recording medium on which a program for executing the above-described inclusion relationship determination and registration is recorded. This program registers the given new code data in the storage means, determines the inclusion relationship between the given new code data and the code data already registered in the storage means, and determines that there is an inclusion relationship. The computer is controlled so that the inclusion relation of the determined code data is associated with the code data and registered in the storage means.

By causing the computer to execute this program, the registration of new code data given and the registration of data indicating the inclusion relationship between the code data and the already registered code data are automatically performed. Done in

[0022]

FIG. 1 is a block diagram showing an electric configuration of a thesaurus generating apparatus.

The thesaurus generation device includes a CPU 1. The CPU 1 has an input device 2 for receiving an input.
(Keyboard, mouse, etc.), an output device 3 (CRT display device, printing device, etc.) for visually outputting the processing results, etc., a program area for storing programs to be executed by the CPU 1, a work area for various operations, etc. Memory 4 for providing area, buffer area, etc., CD-ROM 7
CD-ROM for reading programs and data recorded in
Apparatus 5, floppy (registered trademark) disk (FD) 8
Read programs and data recorded in the FD8
An FD device 6 for recording programs and data and a hard disk device 9 for reading and writing programs and data to and from a hard disk (HD) are connected.

In the CD-ROM 7, a new word is registered in a thesaurus 10 described below, and the upper and lower concepts between words newly registered in the thesaurus 10 and words already registered in the thesaurus 10 are registered. A thesaurus generation program for registering data indicating the relationship. The thesaurus generation program is read by the CD-ROM device 5 and installed on the HD of the HD device 9.

The HD of the HD device 9 has an operating system for totally controlling the entire thesaurus generation device.
The system registers (records) multiple words (terms, keywords) and, for at least some of the registered words, describes the semantic relationship between the word and other registered words. Thesaurus with registered data
10 and the thesaurus generation program described above are stored. The thesaurus generation program stored in the HD is temporarily stored in the memory 4 or expanded.
The thesaurus generation device is controlled by the thesaurus generation program, and a thesaurus generation process described in detail below is performed.

FIG. 2 conceptually shows an example of the data structure of the thesaurus 10 stored in the HD.

The thesaurus 10 includes "clothing" and "shoes".
Seven words of "shoes", "women's shoes", "athletic shoes", "sports shoes" and "star athletic shoes" are registered. In addition, relation data (indicated by a line with an arrow and two alphabetical characters in the vicinity thereof) indicating a semantic relation between a registered word and another registered word is registered.

The relation data includes BT, NT, UF and R
There are four types of T, and the semantic relationship between one registered word and another registered word is indicated by these four types of relation data. BT, NT, UF and R
The meaning of T is as follows.

Broader term (BT): A word having a broader meaning (broader concept) with respect to a certain word. For example, “shoes” has a broader meaning that includes “athletic shoes” (“shoes” encompasses “athletic shoes”), so “shoes” is positioned as a broader term for “athletic shoes” . "Sports shoes" and "shoes" are related by relational data representing the BT (indicated by a line with an arrow starting at "shoes" and ending at "shoes") ("Sports shoes" BT "shoes").

Further, in the thesaurus 10, "shoes" is also a generic term of "woman shoes". “Sports shoes” is a broader term for “stars athletic shoes”, and “shoes” is a broader term for “sports shoes”. According to the relation data representing the BT, “women's shoes” are related to “shoes”, “star athletic shoes” are related to “athletic shoes”, and “sports shoes” are related to “shoes”.

Subordinate term (Narrow Term): A word having a subordinate meaning (narrower concept) with respect to a certain word. For example, "athletic shoes" is a word having a lower conceptual meaning of "shoes", so "exercise shoes" is positioned as a lower conceptual word of "shoes". "Shoes" and "athletic shoes" are related by relation data representing NT ("shoes" NT "exercise shoes").

The upper concept word (BT) and the lower concept word (N
T) is a paired relationship, and the relationship data representing NT and the relationship data representing BT also form a pair. “Shoes” are related to “women's shoes”, “athletic shoes” are related to “star athletic shoes”, and “shoes” are related to “sports shoes” according to the relationship data representing NT.

Synonym (UF: Used For): A word having a meaning (concept of consent) synonymous with a certain word. For example, "shoes" and "shoes" are words having the same meaning. “Shoes” and “shoes” are correlated with each other by relational data representing UF (“shoes” U
F "shoes"). In the thesaurus 10, "exercise shoes" and "sports shoes" are also mutually related by relational data representing UF as words having the same meaning.

Related word (RT): A word having a concept related to a certain word. In the thesaurus 10, "shoes" and "clothing" are related to each other by relation data representing RT as words which are related to each other ("shoes" RT "clothing").

A word registered in the thesaurus 10 is given a semantic relationship with other registered words by the above-described four types of relation data. Of course, a plurality of words may be related to one word by the relation data. In addition, a word having no relation with other registered words may be registered in the thesaurus 10. Generally speaking, in the thesaurus, not only the four types of relational data described above but also other relational data (for example, relational data representing a translation word) are registered.

FIG. 3 is a flowchart of the thesaurus generation processing by the thesaurus generation program.

The thesaurus generation program extracts a word from an input sentence by morphological analysis and, if the extracted word is a word that is not registered in the thesaurus,
The word is registered in the thesaurus 10, and the words newly registered in the thesaurus 10 (hereinafter referred to as newly registered words) are ranked higher among the words already registered in the thesaurus (hereinafter referred to as registered words). If there is a concept word (BT) or a lower concept word (NT), it is associated with the newly registered word by BT and NT.

Using the input device 2 (or from the FD 8)
A sentence is input to the thesaurus generation device. Morphological analysis is performed on the input sentence (step 21). Algorithms for Japanese morphological analysis include “JU” from Kyoto University.
MAN "and" ChaSen "of Nara Institute of Science and Technology are known.

It is assumed that a sentence "cat playing sports while wearing boots" is input. FIG. 4 shows an example of the result of morphological analysis of this sentence. FIG. 5 shows how the thesaurus shown in FIG. 2 is updated (changed) by the processing of the thesaurus generation program when the text "cats playing sports while wearing boots" is input to the thesaurus generation device.

In the morphological analysis process, a Japanese sentence is divided into a plurality of words (words) by the part of speech. The sentence "cats that wear sports with boots" is "boots" (common noun), "wo" (case particle), "hate" (verb) (conjunctive form of "haku"), "sports" (normal Noun), "do"
(Verb) and "cat" (common noun). Only nouns ("boots", "sports", and "cats") are extracted from the divided words.

For each of the extracted words (nouns), it is determined whether the word has already been registered in the thesaurus 10 (step 22). This judgment is based on the character code representing the extracted word (for example, ASCII code)
This is performed by comparing the character codes of the registered words registered in the thesaurus 10 with those of the registered words.

If the extracted word is already registered in the thesaurus 10 (YES in step 22), the next word extracted by the morphological analysis process (step 21) is registered in the thesaurus. Is determined (NO in step 23, step 22).

If the extracted words are words that have not been registered in the thesaurus 10 (N in step 22)
O), the extracted words are indexed and newly registered in the thesaurus 10 (step 24) (newly registered words).

The newly registered word is already in the thesaurus
The inclusion relation with the word registered in 10 is checked. That is, one of the registered words already registered in the thesaurus 10 is selected, and a higher concept word (BT) or a lower concept word (NT) is selected between the selected registered word and the newly registered word. It is determined whether there is a relationship (steps 26 and 28).

First, with respect to a newly registered word, it is determined whether or not the selected registered word is a broader concept word (BT) (step 26). This determination is made by comparing the character code representing the newly registered word with the character code representing the selected registered word, and determining whether the character code representing the newly registered word includes the character code representing the registered word.

For example, the character code representing the newly registered word "boots" includes the character code representing the registered word "shoes" in the thesaurus 10 (matching in the "shoes" portion). For the newly registered word "boots", the registered word "shoes"
Is a broader term ("shoes" includes "boots") (YES in step 26). Thesaurus 10
, Relation data (BT) indicating that "shoes" is a broader term of "boots" is registered (step 27, FIG. 5). Further, since BT and NT form a pair, relation data (NT) indicating that "boots" is a lower concept word of "shoes" is also registered.

If the selected registered word is not a broader concept word (BT) for the newly registered word (step 26).
NO), it is determined whether the selected registered word is a lower concept word (NT) (step 28).

When a character code representing a newly registered word is included in a character code representing a registered word (for example, a character code representing a newly registered word “sports” is a character code representing a registered word “sports shoes”) ), The selected registered word is determined to be a lower concept word (YES in step 28). In the thesaurus 10, relation data (NT) indicating that the registered word ("sports shoes") is a lower concept word of the newly registered word ("sports") is registered, and the newly registered word ("sports") is registered. Relationship data (B indicating that it is a broader concept word of the registered word ("sports shoes")
T) is registered (step 29, FIG. 5).

The newly registered word is compared with all the registered words registered in the thesaurus 10 (or the registered words within a specific range) (NO in step 30, NO in step 30).
Five).

In the case where all the registered words are neither the superordinate concept word nor the subordinate concept word with respect to the newly registered word (step 26)
, NO in step 28), and the newly registered word is registered in the thesaurus 10 without having a relationship between the already registered word and NT and BT (for example, “cat” in FIG. 5).

When it is determined whether or not there is a relation between a higher concept word or a lower concept word between the newly registered word and all the registered words (YES in step 30), the extracted words are extracted in morphological analysis (step 21). It is determined whether or not the next word is registered (step 22). If the word has not been registered yet, after registration in the thesaurus 10 (step 24), the process proceeds to the determination of the relation between the upper concept word or the lower concept word (steps 25 to 30).

When the above processing (steps 22 to 30) is completed for all the extracted words, the thesaurus generation processing ends (YES in step 23).

As described above, in the thesaurus generation processing, words that have not yet been registered in the thesaurus 10 included in the input sentence are registered in the thesaurus 10, and furthermore, the relation indicating the higher concept word (BT) and the lower concept word (NT) Data is automatically registered. The user registers the related data indicating the synonym (UF) and the related word (RT) using the input device 2. Of course, BT and N
The user can also register the relation data indicating T using the input device 2.

The generated thesaurus can be used for document search processing. For example, when searching for (extracting) a document containing the word “exercise shoes” (key word) from the documents stored in the database, the thesaurus 10 can be used to retrieve a document containing “athlete shoes”. Not only can you search for, but also "exercise shoes"
Words registered in association with (for example,
"Shoes" or "Women's shoes" which are further related via "Shoes")
Are also searched for as documents containing similar key words.

In the above-described embodiment, a word (Japanese) is registered in the thesaurus 10, and the character code representing the newly registered word and the character code representing the registered word are compared, so that the newly registered word and the registered word are compared. (The relation between the broader concept word (BT) or the lower concept word (NT)) is determined. If the thesaurus 10 has registered data on chemical formulas, for example, the newly registered chemical formula (or a specific part included in the chemical formula) is included in the already registered chemical formulas, or The inclusion relation (superordinate concept or subordinate concept) may be determined based on whether or not the included chemical formula is included. As described above, in the determination of the superordinate concept or the subordinate concept, the element of the determination may be determined according to the type of data registered in the thesaurus 10.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an electrical configuration of a thesaurus generation device.

FIG. 2 conceptually shows a data structure of a thesaurus.

FIG. 3 is a flowchart illustrating processing of a thesaurus generation program.

FIG. 4 illustrates an example of a morphological analysis process.

FIG. 5 conceptually shows a data structure of a thesaurus.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 CPU 2 Input device 3 Output device 4 Memory 5 CD-ROM device 6 FD device 7 CD-ROM 8 FD 9 Hard disk device 10 Thesaurus

Claims (5)

[Claims] 1. A storage means for storing a thesaurus describing a relationship between a registered word and at least a part of the registered word, the relation between the word and another word, an input means for inputting a sentence, Word extraction means for extracting words contained in the text input by the input means,
Registration determining means for determining whether the word extracted by the word extracting means is a word registered in the thesaurus, and determining that the extracted word is not a word registered in the thesaurus by the registration determining means In this case, the registration means for newly registering the extracted word in the thesaurus, comparing the character string of the newly registered word newly registered in the thesaurus with the character string of the already registered word already registered in the thesaurus, and determining their inclusion relations Means for determining inclusion relation,
And when the inclusion relation determining means determines that the character string of the newly registered word includes the character string of the registered word, the relation that the registered word is a superordinate concept word of the newly registered word, The relation that the newly registered word is a lower concept word of the above-mentioned registered word is registered in the thesaurus, and when it is determined that the character string of the newly registered word is included in the character string of the already-registered word, the above-mentioned word is registered. A thesaurus generation device comprising: a relationship registration unit that registers in a thesaurus a relationship that a registered word is a lower concept word and a relationship that a newly registered word is a higher concept word of the registered word. 2. A method of generating a thesaurus describing a relationship between a registered word and at least a part of the registered word, and a relationship between the word and another word, wherein the thesaurus is included in a given sentence. The extracted word is extracted, and it is determined whether the extracted word is a word registered in the thesaurus. If it is determined that the extracted word is a word not registered in the thesaurus, the extracted word is determined. Is newly registered in the thesaurus, the character string of the newly registered word newly registered in the thesaurus is compared with the character string of the registered word already registered in the thesaurus, their inclusion relation is determined, and the new registration is performed. If it is determined that the word character string includes the registered word character string, the relation that the registered word is a superordinate concept word of the newly registered word and the new registered word is When the relation that the word is a lower concept word of the registered word is registered in the thesaurus, and it is determined that the character string of the newly registered word is included in the character string of the registered word, the registered word is replaced with the new registered word. Is a lower concept word and a relation that the newly registered word is a higher concept word of the already registered word is registered in a thesaurus. 3. A program for creating a thesaurus describing a relationship between a registered word and at least a part of the registered word, the relationship between the word and another word, wherein the given word is , Determine whether the word is registered in the thesaurus, and if it is determined that the word is not a word registered in the thesaurus, newly register the word in the thesaurus and newly register in the thesaurus The new registered word character string is compared with the registered word character string already registered in the thesaurus to determine their inclusion, and the newly registered word character string is replaced with the registered word character string. If it is determined that the word is included, the relation that the registered word is a superordinate term of the newly registered word and the relation that the newly registered word is a lower concept word of the registered word Is registered in the thesaurus, and when it is determined that the character string of the newly registered word is included in the character string of the registered word, the relationship between the registered word and the new A recording medium recording a thesaurus generation program for controlling a computer to register a relation that a registered word is a broader concept word of the registered word in the thesaurus. 4. A storage means for storing code data and an inclusive relation between code data, a means for inputting new code data, and a code for registering new code data input by the input means in the storage means. Data registration means, means for determining the inclusion relationship between newly input code data and code data already registered in the storage means, and the inclusion relation of the code data determined to have an inclusion relation to those codes. An inclusive relation registering means for registering in the storage means in association with data. 5. The new code data given is registered in the storage means, the inclusion relation between the given new code data and the code data already registered in the storage means is determined, and the inclusion relation is found. A recording medium storing a program for controlling a computer so that the inclusion relation of the code data determined as above is associated with the code data and registered in the storage means.