JP2001014348A - Computer sharing technology dictionary - Google Patents

Abstract

PROBLEM TO BE SOLVED: To store the basic knowledge and relative knowledge of technical terms on computers, to enable the computers to transfer or process the stored knowledge, and to optionally take out them for use. SOLUTION: The basic knowledge and relative knowledge of technical terms represented by a one-story predicate logical expression are stored while classified into three kinds of finite concept elements structured by use of a combination of a relative concept link and term nodes 2, 5, and 8. The basic knowledge of index words 1, 4, and 7 is stored in a star type concept element (a), the relative knowledge is stored in a tree type concept element (b), and associate knowledge is stored in a triangular concept element (c) respectively; when there is an inquiry about a specific index word, secondary derivative knowledge is created and provided in addition to the basic knowledge and relative knowledge in the original form. A document having no term limitation or a term of a knowledge base is replaced with a standard word, i.e., index word by means of a homonym standardizing function to improve the precision and the processing speed of knowledge processing.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technology for sharing and reusing basic knowledge and background knowledge of technical terms used for expressing explicit knowledge between humans and computers.

[0002]

2. Description of the Related Art Human knowledge is shared between humans as knowledge specified using characters, symbols, figures, and drawings, that is, document knowledge. There are various technical term dictionaries as a means for surely sharing knowledge. In the conventional technical term dictionary, basic knowledge and background knowledge of terms are described in natural language, so that knowledge is stored in computers, exchanged and processed between computers, and humans arbitrarily take out and reuse from computers It is very difficult to realize knowledge sharing between humans and computers as much as possible. The science and technology thesaurus contains the parent-child relationships and related terms of terms required for literature search.
The basic knowledge of the term is not stored and the background knowledge is partial. The term "storage" means to store and store terms and concepts in a predetermined concept structure. The same applies hereinafter.

[0003] Although some of the technical term dictionaries and thesauruses are digitized and can be used on computers, they are simply characters, symbols, and figures converted into digital information, and computers share basic knowledge and background knowledge of terms. It does not provide knowledge reproduction or inference processing.

[0004] The thesaurus shows a simple parent-child relationship of terms, but does not have an explicit conceptual structure or ontology that guarantees the consensus of engineers. Technical documents generated by many engineers in cooperation often use different terms for the same knowledge. If the expressions are different, direct comparison and collation by computer are impossible.

[0005] There is a frame as a knowledge representation method.
Frames are effective for class definition and instance representation of mechanisms and mechanical parts. However, in terms of the definition of technical terms used to define frames and the expression of background knowledge, the multifunctionality of frames is rather an obstacle to ontology development aimed at structuring and standardizing a unified term concept. It is difficult and impractical to cover the entire knowledge space with one expression.

[0006]

According to the conventional method, it is very difficult to store basic knowledge and related knowledge of terms in a computer, exchange and process the data with each other, and take out and use the data from the computer at will. is there.

[0007] In order to store document knowledge represented by characters, symbols, figures and tables in a computer as a set of first-order predicate logical expressions, a unified means for storing basic knowledge of terms and related knowledge in a computer is required. . Basic knowledge can be expressed by a combination of a finite number of predicates and terms.Relational knowledge consists of hierarchical knowledge of terms connected by partial whole relationships, class relationships, causal relationships, etc., and a network of terms connected by associative relationships. There is associative knowledge expressed.

In order to computer-process document knowledge without any term restrictions, it is necessary to unify various synonyms into standard words and express knowledge using standard words. Standard words maintain unity, and compound words are expressed as ordered pairs of standard words.

It is necessary to develop an algorithm for arbitrarily extracting the knowledge stored in the computer according to the purpose of use. In particular, the algorithm for automatically generating search schemes necessary for extracting target areas in document processing, such as collating technical documents, editing new documents from existing documents, applying rule knowledge, and creating instances using class libraries, as well as storing knowledge. We need an algorithm to create the derived knowledge used.

[0010]

Means for Solving the Problems A first configuration of the present invention is a finite concept element. Knowledge is stored for each headword using the following three types of finite concept elements. This makes it possible for both humans and computers to reproduce and use them together. The finite concept element is a concept element that combines a link that stores a related concept of an entry word and a node that stores a term to be related, and has three elements of a star-shaped concept element, a tree-shaped concept element, and a triangle concept element. The headword has a star concept element that stores basic knowledge and a triangle concept element that stores associative knowledge as needed. Among the headwords, headwords that form a term set combined by a specific relational concept such as a partial whole relationship are referred to as whole words, and are head words of tree-type concept elements. FIG. 1 shows a structural diagram of the finite concept element. (A) is a star-shaped conceptual element, (b) is a tree-shaped conceptual element, and (c) is a triangular conceptual element.

FIG. 1A shows a star-shaped conceptual element. The star concept element is used to store the basic knowledge of the headword. Stores 1: N term relationships. Usually, the concept recalled by a human from one headword is about 6 to 10, and the number of nodes of the star-shaped concept element can be practically limited to 6 to 10. The number of nodes is unified in the same dictionary.

FIG. 1B shows a tree-shaped concept element. The tree-shaped concept element stores a hierarchical finite concept set in which relational concepts such as partial whole relations, class relations, and causal relations are connected as links. The head word of the tree is the whole word of the concept set. Isolated words and straight hierarchies are tree variants. The number of branches of each node in the tree shape is arbitrary.

FIG. 1C shows a triangular concept element. The triangular concept element stores a finite concept set in which different terms in the knowledge domain are connected by using an associative relationship as a link. Reference numeral 7 stores the headword and the associative concept in two nodes of reference numeral 8. If there is a concept associated with the concept of any two nodes, a new node is arranged to expand the network of the associative relationship.

The second configuration of the present invention is a function for standardizing synonyms. A synonym link is provided for one link of the star concept element, and a synonym node stores free words that are synonymous with the headword. All documents or knowledge base terms expressed by logical expressions processed by a computer search for free words of synonym nodes and replace them with corresponding headwords.

A third configuration of the present invention is an algorithm for automatically generating a search scheme for corresponding knowledge. Document processing, which is the object of the present invention, is knowledge processing between two documents or knowledge bases, and a search scheme for improving the extraction accuracy of a processing target part is automatically generated from stored knowledge using structural characteristics of conceptual elements. Generated.

A fourth configuration of the present invention is an algorithm for generating secondary derivative knowledge based on the principle of fusion of first-order predicate logic. Derived knowledge is obtained from the logical formula stored in the star concept element,
Also, related knowledge is derived from the tree-shaped concept element and the triangle concept element. The use of derivative knowledge can improve the quality of document processing.

Although not described in the specification, it is possible to create tertiary derived knowledge from basic knowledge using inference rules specific to the headword.

[0018]

The computer shared dictionary according to the present invention stores entry words, their entire words and associated words. The storage and regeneration of these terms and knowledge will be described. FIG. 1 shows the structure of three types of finite concept elements storing knowledge, and FIG. 2 shows the storing and reproducing operations of each of the knowledge. 1A shows a star concept element, FIG. 1B shows a tree concept element, and FIG. 1C shows a triangle concept element. Triangle elements work similarly for quadrilateral elements.

FIG. 2A shows the storing and reproducing operation of the knowledge of the star-shaped concept element. FIG. 2A shows a link set Ri (i
= 1 to 6) and the connection node stores basic knowledge unique to the headword. Knowledge is a logical expression Ri (D, T (R
i)), the entry word D is stored in the code 1, the predicate Ri is stored in the link set, and the term T (Ri) connected by Ri is stored in the corresponding node. By structuring headword concepts using concept elements, humans can visually recognize, edit, and update the concept structure, and computers can read logical formulas mechanically and perform necessary knowledge processing. Will be possible. The number of link sets can be selected arbitrarily, but within the same dictionary, the number and predicate set are fixed, and secondary
Generate secondary derivative knowledge.

FIG. 2B shows the storing and reproducing operation of the knowledge of the tree concept element. Stores hierarchical knowledge composed of hierarchical headword sets connected by specific relational concepts.
The headword at the top of the headword set is defined as a whole word that is a marker of hierarchical knowledge. The hierarchical knowledge of the whole word C is expressed in the form of Rh (C, Tj, Tk) using the relation concept Rh. The headwords Tj and Tk are headword sets connected by the relation concept Rh, and the number of layers and the size of each layer can be arbitrarily increased or decreased. Reproducible knowledge is hierarchical knowledge that is linked by basic knowledge of headwords C, Tj, and Tk and their related concept Rh.

FIG. 2C shows the storing and reproducing operations of the knowledge of the triangular concept elements. Associative relationship Ra of any headword Ai
Are the associated words Ej and Fk. Ai, Ej,
Fk is linked by Ra. Renewable knowledge is Ej,
This is the basic knowledge of Fk and the knowledge representing their association.

FIG. 8 shows an algorithm for converting an arbitrary term or synonym into a headword. When a term to be converted is input, a search is first made for a headword in the entire dictionary. If it is not a headword, the synonym node is searched similarly, and if there is a corresponding term in the synonym node, the headword is output. If there is no synonym node, a warning is displayed indicating that the dictionary is not registered. Register new synonyms according to the warning.

FIG. 9 shows an algorithm for generating a search scheme suitable for a search purpose from a headword related to an arbitrary headword. When the headword is input, the related headword and associative word of the whole word are searched, and if there is a corresponding word, the keyword is input to the keyword filter 73 together with the headword to remove a keyword that does not match the search purpose, and a search scheme is performed. Get the set of keywords that make up.

FIG. 10 shows an algorithm for creating secondary derivative knowledge from an arbitrary headword and a headword related to the headword. When the headword is input, the related headword and associative word of the whole word are searched, and the relevant link set and the inference rule specific to the headword are selected from the link set / inference rule storage device 82 and executed. Output secondary and tertiary derivative knowledge. The link set is unified for each classification based on the nature of the headword such as the concept of an object, the concept of a class, and the concept of an event. The inference rules include the rules of the link set itself, inference rules based on hierarchical knowledge and associative knowledge, and inference rules based on basic knowledge of headwords.

[0025]

Embodiments of the present invention will be described with reference to FIGS.
5 and 6. FIG. 3 shows a state in which the basic knowledge of the headword [gear] is stored in a star-shaped concept element. FIG. 4 shows a state in which hierarchical knowledge of the whole word [gear device] of [gear] is stored in a tree-shaped concept element. FIG. 5 shows a state in which the associative knowledge of the entry word [gear] is stored in the triangular concept element. FIG. 6 shows a state in which basic knowledge of the entire term [gear device] is stored.

FIG. 7A is an expression of the stored knowledge and the reproduced knowledge of the entry word [gear] in FIG. 3 by a logical expression. Regenerated knowledge creates secondary derivative knowledge in addition to the stored hierarchical primary knowledge.

FIG. 7B is an expression of the stored knowledge and the reproduced knowledge of the general term [gear unit] in FIG. 4 by a logical expression. Regeneration knowledge creates secondary derivative knowledge in addition to the stored primary knowledge.

FIG. 7C is an expression of the stored knowledge and the reproduced knowledge of the entry word [gear] in FIG. 5 by a logical expression. Regeneration knowledge creates secondary derivative knowledge in addition to the stored primary knowledge. Regeneration knowledge is background knowledge for the basic knowledge of the headword [gear].

[0029]

According to the first aspect of the present invention, basic knowledge and related knowledge of a headword can be stored in a state where both a computer and a human can perform knowledge processing. Further, derived knowledge, which is background knowledge of the headword, can be created from the knowledge stored in the computer by an inference rule using a knowledge storage structure. The present invention can be applied to legal documents, medical documents, and the like.

According to the second aspect of the present invention, various technical terms used in a technical document or a knowledge base can be mechanically converted into headwords having definitions of basic knowledge and related knowledge. According to the present invention, the term space of the dictionary is greatly reduced, and high-speed and efficient knowledge processing can be performed.

According to the third aspect of the present invention, a technical document,
A search scheme required for searching various information such as a knowledge base and a logical expression can be generated mechanically. Since a search scheme suitable for the search purpose can be constructed, the speed of knowledge processing can be increased.

According to the fourth aspect of the present invention, the secondary,
Can create tertiary derivative knowledge. The accuracy of knowledge processing such as collation and synthesis of technical documents, construction of a knowledge base, and construction of a frame library is improved.

[0033]

[Brief description of the drawings]

FIG. 1 is a finite concept element used in the implementation of the present invention. (A) is a star-shaped conceptual element, (b) is a tree-shaped conceptual element, and (c) is a triangular conceptual element.

FIG. 2 is an explanatory diagram of a function of storing and reproducing knowledge of a finite concept element according to the embodiment of the present invention. (A) is a star-shaped conceptual element, (b) is a tree-shaped conceptual element, and (c) is a triangular conceptual element.

FIG. 3 is an example of a headword [gear] of a star-shaped concept element used in the embodiment of the present invention.

FIG. 4 is an example of a general term [gear device] of a tree-shaped concept element used in the embodiment of the present invention.

FIG. 5 is an example of a headword [gear] of a triangular concept element used in the embodiment of the present invention.

6 is a general term [gear device] of the headword [gear] in FIG.
It is a star-shaped concept element that stores basic knowledge of.

FIG. 7 is a logical expression of the knowledge stored in the conceptual elements of FIGS.

FIG. 8 is an algorithm for converting an arbitrary term used in the embodiment of the present invention into a headword.

FIG. 9 is an algorithm for generating a search scheme for searching related to any headword used in the embodiment of the present invention.

FIG. 10 is an algorithm for creating secondary derivative knowledge of any headword used in the embodiment of the present invention.

[Explanation of symbols]

1, 4, 7 headword 2, 5, 8 node 3 link set 6, 9 link 10, 13, 16 stored knowledge 11, 14, 17 place to store knowledge 12, 15, 18 logical expression of reconstructed knowledge 19 ２４24 Terms constituting basic knowledge of entry word [gear] 25-30 Link set constituting basic knowledge of entry word [gear] 25 Entire word link 26 Synonym link 27 Function link 28 Behavior link 29 Application link 30 Feature link 31 Entire word 32 to 35 Headword 36 Relational concept link 37 Headword 38, 39 Headword associated with headword 37 40 Associative relation link 41 to 46 Terms constituting basic knowledge 41 to 46 Heading Term node of the word [gear device] 47-52 Link set of the entry [gear device] 53-55 Collection knowledge of FIG. 3 56-58 Reproduction knowledge of FIG. Processing elements of the processing elements 76 to 86 derived knowledge creation algorithm of the processing elements 67 to 75 search scheme generation algorithm word conversion algorithm

Claims (4)

[Claims] 1. The basic knowledge and related knowledge of an entry word are stored in at least one or more finite concept elements structured using a combination of a related concept link and a term node, and optionally, the stored knowledge and derived knowledge are stored. A computer sharing technology dictionary characterized by having the function of reproducing. 2. A set of links and nodes of a star-shaped concept element structured using a combination of relation concept links and term nodes that store basic knowledge and related knowledge of headwords are synonymous with synonym links. A computer shared technology dictionary having word nodes and having a function of converting synonyms stored in synonym nodes into headwords. 3. A function of storing basic knowledge and related knowledge of a headword in a finite concept element structured using a combination of a related concept link and a term node, and having a function of optionally reproducing the stored knowledge. An algorithm for automatically generating a search scheme for searching for corresponding knowledge of a partner for an arbitrary headword in a contrast process between two knowledge bases using a computer-usage technology dictionary as a feature. 4. A function of storing basic knowledge and related knowledge of a headword in a finite concept element structured using a combination of a related concept link and a term node, and optionally reproducing the stored knowledge. An algorithm that infers and creates additional knowledge using the knowledge stored in the computer-usage technology dictionary.