JPH04372016A - Japanese language understanding device - Google Patents

Abstract

PURPOSE:To provide a Japanese language understanding device for development, expansion, and preservation of a system in such a manner that a manual is prepared by natural word is Japanese language even if a user do not have any knowledge of a program. CONSTITUTION:A knowledge compiler part 2 analysis the inputted Japanese knowledge to be converted into a C language program. A knowledge acquisition part 1 expands a knowledge compiler part 2 by inputting the natural word in Japanese language and acquires the knowledge of the Japanese language to be the input of the knowledge compiler part 2. A screen input part 6 inputs atom data which is one of the Japanese language knowledge for the application execution. Thus, the operation execution program is prepared by inputting natural word in Japanese language.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a Japanese language understanding device for developing, expanding, and maintaining a system using Japanese natural language.

0002

2. Description of the Related Art The current situation in the world of software is that the supply of software is not keeping up with the demand for software, which is called a software crisis. The cause of this is not only assembler language, but also C, COBOL, FORTRAN.
General high-level programming languages such as these are also difficult for general users, and it is difficult to develop them in parallel with work.Currently, even work that is closely related to the field is developed by outsourcing to software engineers. It is said that progress is being made.

In conventional business development that requires system development, for example, as shown in FIG. 11, business is first analyzed (step S11) and a business plan is created (step S12). Then, create a proposal based on it (step S1
3) Determine the content of business development, request the content to the system (step S14), operate it (step S15),
If there is a bug fix, function change, function expansion, etc., issue a change request (step S16), and issue a system request again (step S16).
14) I will do it.

[0004] When a system is requested, development proceeds as shown in FIG. 12 in accordance with the conventional system development procedure. That is, first, a system analysis is performed (step S21),
A system plan is made (step S22). Next, the system configuration is determined by system design (step S23),
A program is determined by program design (step S24)
), after actually coding the program (step S25) and testing (step S26), operate and maintain the system (step S27), and if there are any bug fixes, function changes, or enhancements, update the content. You will move on to each step and modify the program accordingly.

[0005]

[Problem to be Solved by the Invention] However, in a development system in which business experts, system developers, and maintainers are separated as described above, the effort of analysis, planning, design, documentation, and maintenance overlaps. In addition, from the perspective of business experts, it is difficult to understand how systems are developed, and from the perspective of system developers, it is difficult to grasp the know-how of operational experts, so it is difficult to understand how systems are developed. Bugs are more likely to occur, and bug fixes, feature changes, and feature enhancements may not be handled well. Furthermore, the know-how of business experts is not preserved in a systemized document.

Case tool environment technology that brings software engineering techniques to system development (Special feature on "Information Processing": Case Environment, VOL. 31, No. 8, 1990)
was researched and developed, and its commercialization (``Nikkei Computer'' 1)
989.2.13, pp66-84) has been advanced, and Japanese 4GL for programming in Japanese (``Nikkei Computer'' 1988.9.12, pp89-96) has been advanced.
) has been commercialized, but it is still in the stage of being used as a support tool for software engineers, and it is difficult for general users to use it.

[0007] From the perspective of usage by general users,
HyperCard, a hypertext system
(Apple Inc., 1987) is a scripting language called HyperTalk, but its use is limited to presentations. As an approach for general users, technology for analyzing natural language (Special feature on "Information processing": Natural language understanding, VOL30, No.
10, 1989) and its Japanese language processing (Nagao: ``Japanese Information Processing'', Institute of Electronics and Information Engineers Publishing, 1984) have been researched and developed, and recently there has been active research into context understanding and discourse understanding models. However, due to the complexity of language,
The more we try to understand it, the more complex the algorithm becomes, and the memory capacity and speed are such that it cannot be put to practical use at the personal computer level. In actual use, there are lex, a lexical analysis generator that generates a finite automaton state transition table and its driver system from regular expressions and their extensions, and yac, an upward syntactic analysis generator, that run on UNIX.
c, etc., but it is strongly intended for computer languages, and is not suitable for generating Japanese compilers for business purposes.

[0008] Furthermore, research on the above-mentioned context understanding and discourse understanding has just begun, and it will take time to put them into practical use. As described above, at present, the development system in which business experts, system developers, and maintainers are separate has not been resolved, resulting in the above-mentioned problems. The present invention has been made in view of the above circumstances, and is a Japanese system that allows general users to develop, expand, and maintain systems in the same manner as creating manuals in Japanese natural language without having any knowledge of programming. The purpose is to provide a word understanding device.

[0009]

[Means for Solving the Problems] The present invention includes a knowledge compiler unit that analyzes input Japanese knowledge and converts it into a C language program, and a knowledge compiler unit that analyzes input Japanese language knowledge and converts it into a C language program, and A knowledge acquisition section that acquires Japanese knowledge that will be expanded and input to the knowledge compiler section, and a screen input section that inputs atom data, which is one of the Japanese knowledge for application execution, are provided. The system is characterized by a structure in which a business execution program is created by inputting natural language.

0010

[Operation] The knowledge compiler section analyzes the input Japanese knowledge and converts it into a C language program. The knowledge acquisition section expands the knowledge compiler section by inputting Japanese natural language, and acquires Japanese knowledge that is input to the knowledge compiler section. The screen input section inputs atom data, which is one of Japanese knowledge, for application execution. Therefore, a business execution program is created by inputting Japanese natural language.

[0011]

Embodiments Hereinafter, embodiments of the present invention will be explained in detail with reference to the drawings. FIG. 1 is a block diagram of a Japanese language understanding device according to an embodiment of the present invention, which includes a knowledge acquisition section 1, a knowledge compiler section 2, a basic library section 3, and a C language compiler and linker. It is composed of a section 4, a knowledge processing execution section 5, and a screen input section 6. The knowledge acquisition unit 1 and the screen input unit 6 are configured by, for example, a keyboard device, a display device, and the like. Japanese knowledge is input to the knowledge acquisition unit 1 by a business expert using Japanese natural language. Knowledge compiler part 2
generates a C program source corresponding to the Japanese knowledge from the knowledge acquisition unit 1. The C language compiler and linker section 4 compiles the C program source from the knowledge compiler section 2 and links it with the basic library from the basic library section 3. The user inputs into the screen input section 6 a condition file, which is the content of changes to one atom data of Japanese knowledge, in accordance with the execution conditions. The knowledge processing execution unit 5 executes the business system based on data from the C language compiler and linker unit 4 and the screen input unit 6.

FIG. 2 is a block diagram of the knowledge acquisition unit 1. The knowledge acquisition unit 1 includes a meaning table registration unit 11, a predicate table registration unit 12, a frame table registration unit 13, an object data input unit 14, and an atom data input unit 14. Table registration section 15
and a frame data input section 16. The meaning table registration section 11 is for registering a plurality of expressions having the same meaning or expressions unique to the user in the meaning table of the meaning table section 17. The predicate table registration unit 12 stores a predicate table listing function names, function names, and arguments of new functions and existing systems in the predicate table unit 1.
This is for registering in the predicate table of No. 8. The frame table registration unit 13 stores a frame table defining the framework of the user database in the frame table unit 19.
This is for registering. The object data input section 14 is for inputting object data, which is the main body of Japanese knowledge, and is the execution content based on actual Japanese knowledge. The atom table registration section 15 is for registering an atom table, which is a Japanese variable having a value, type, and state, in the atom table section 21. The frame data input section 16 is for inputting frame data, which is data of the frame table of the frame table section 19. The meaning table registration section 11, predicate table registration section 12, and frame table registration section 13 are for extending the knowledge compiler section 2, and the object data input section 14, atom table registration section 15, and frame data input section 16 is for acquiring knowledge that is input to the knowledge compiler section 2.

FIG. 3 is a block diagram of the language analysis section of the knowledge compiler section 2. The language analysis section of the knowledge compiler section 2 includes a lexical analysis section 31, a syntactic analysis section 32, a semantic analysis section 33, and a semantic table. 17, a predicate table section 18, a C standard function section 34, an atom table section 21, and a frame table section 19. The lexical analysis unit 31 performs lexical analysis of the object data input to the object data input unit 14 of the knowledge acquisition unit 1 by referring to the meaning table in the meaning table unit 17 . The syntactic analysis unit 32 refers to the meaning table in the meaning table unit 17 and performs syntactic analysis based on the results analyzed by the lexical analysis unit 31. The semantic analysis unit 33 refers to the predicate table in the predicate table unit 18, the C standard function in the C standard function unit 34, the atom table in the atom table unit 21, and the frame table in the frame table unit 19, and
The simple sentence extracted by the syntactic analysis unit 32 is semantically analyzed.

FIG. 4 is a block diagram of the basic library section 3, which includes a C language related library section 41 and a knowledge related library section 4.
2 and a user library section 43. The C language related library of the C language related library section 41 performs program processing, message output, and error processing. The knowledge-related library of the knowledge-related library section 42 performs knowledge processing of atoms and frames. The user library of the user library section 43 registers system-provided general-purpose functions and user-specific functions.

FIG. 5 is an explanatory diagram of the Japanese knowledge obtained by the knowledge acquisition unit 1. FIG. 6 is an explanatory diagram of the object data input to the object data input unit 14 of the knowledge acquisition unit 1. It is written using delimiters, and each sentence is separated by ".". Each functional module is called an object. Objects are managed hierarchically, and it is possible to call lower-level functions, or register a group of functions and call them in each part.

FIG. 7 is an explanatory diagram of the meaning table of the meaning table section 17. FIG. 8 is an explanatory diagram of the frame table of the frame table section 19. Next, the operation will be explained. The specialization of work is the meaning table registration unit 11 of the knowledge acquisition unit 1.
The meaning table is registered in the meaning table section 17, the predicate table registration section 12 registers the predicate table in the predicate table section 18, and the frame table registration section 13 registers the frame table in the frame table section 19. This expands the knowledge compiler section 2. Further, when the business is specialized, the atom table registration unit 15 of the knowledge acquisition unit 1 registers an atom table in the atom table unit 21 and inputs frame data to the frame data input unit 16.
stored in the frame table. When the business specialization inputs object data to the object data input section 14 of the knowledge acquisition section 1, the object data is supplied to the knowledge compiler section 2.

When the knowledge compiler unit 2 is supplied with object data from the knowledge acquisition unit 1, as a pre-scanning process, it first cuts out the object and extracts each calling relationship. Next, cut out a sentence for each object. As a result, the lexical analysis section 31 of the knowledge compiler section 2 refers to the meaning table of the meaning table section 17 and assigns meaning to each punctuation character string constituting one sentence of the object.

The syntactic analysis unit 32 maps the relationships of the information from the lexical analysis unit 31 in advance through pre-scanning, extracts conditional sentences, complex sentence structures, parenthetical sentences, and repeated sentences, and in this state, performs the following operations at any time. By passing information to the semantic analysis unit 33,
Stop the divergence of parse tree generation. Therefore, the semantic analysis unit 33
Receive partial simple sentences regardless of sentence structure. For example, in the case of the sentence ``If the limit amount is greater than the limit, the limit value is 3 * limit amount, and the process is interrupted when the limit is exceeded.''
is greater than or equal to the limit" and the conditional form "is the limit
is passed in the standard form ``3*A'' and ``processing is interrupted when the limit is exceeded''.

Operations on knowledge values such as the atom table in the atom table section 21 and the frame table in the frame table section 19 are extracted as noun phrases and written out as needed. For example, "limit amount" can be set using {GetRAtom(
111/*limited amount*/) }. Also, object calling relationships, input/output, four arithmetic operations,
C standard functions of the C standard function section 34, such as value manipulation, are also written out at any time.

For the functions registered in the predicate table of the predicate table section 18, <function name→function name, factor 1
, factor 2, --> is expanded into a predicate logical expression of the form, and is expanded into a function. For example, "processing interrupted due to limit exceeded" becomes <processing interrupted → processing interrupted, limit exceeded>,
Expands into a function. Functions that cannot be extracted by the above operations are treated as errors.

[0021] Thus, the knowledge compiler unit 2 converts the Japanese knowledge from the knowledge acquisition unit 1 into a C
Exported to program source. The C language compiler and linker section 4 compiles the C program source from the knowledge compiler section 2 and links it with the basic library of the basic library section 3 to create a program for executing a business and performs knowledge processing. It is stored in the execution unit 5.

Thereafter, the user inputs each condition factor from the screen input section 6 to the knowledge processing execution section 5 as a condition file that is the change content of the atom data, so that the knowledge processing execution section 5 executes the task. Ru. In this way, a program for executing a business system can be created simply by inputting Japanese knowledge into the knowledge acquisition unit 1 using Japanese natural language. Analyze (step S1), create a business plan (step S2), create a proposal based on it (step S3), determine business development content, and create a manual based on the content, that is, send to knowledge acquisition section 1. Japanese language knowledge is input (step S4), operation and maintenance are performed (step S5), and if bugs are fixed, functions are changed, functions are expanded, etc., a manual is created again (step S4). In other words, business development is the result of system development, the business analysis in step S1 corresponds to system analysis, the business plan in step S2 corresponds to system planning, and the proposal creation in step S3 corresponds to system analysis. This corresponds to design, and the manual creation in step S4 corresponds to program design.

[0023] As a result, a general user who is a business expert can develop, expand, and maintain the system in the same way as creating a manual in Japanese natural language, even if he or she has no knowledge of programming. Furthermore, as in this embodiment, when inputting object data to the knowledge acquisition unit 1, the efficiency of lexical analysis by the lexical analysis unit 31 of the knowledge compiler unit 2 is improved by not allowing solid writing, which is considered difficult to analyze in Japanese. The lexical analysis unit 31 also refers to the meaning table in the meaning table unit 17 in which conjunctions, particles, operators, etc. are registered, and extracts conditions, parentheses, loop relationships, etc. in advance, and the syntactic analysis unit 32 extracts, If the structure is configured to improve syntactic analysis efficiency by performing segmentation processing at any time while maintaining the nested structure based on information from the lexical analysis unit 31, use at the personal computer level becomes easily possible.

Furthermore, as in this embodiment, the meaning table section 1
If multiple expressions with the same meaning and user-specific expressions can be registered in the meaning table 7, and new functions and existing systems can be registered in the predicate table of the predicate table section 18, the language analysis framework can be improved. Extend and generalize
Can support user-specific applications. Further, as in this embodiment, a user's own data format can be registered in the frame table of the frame table section 19, and data contents conforming to the format of the frame table can be input from the frame data input section 16 of the knowledge acquisition section 1 into the frame table. The object data input unit 1 of the knowledge acquisition unit 1 can be registered in the frame table of the table unit 19.
It hierarchically manages the Japanese input from 4, and allows lower-level functions to be called within the Japanese language knowledge, or certain grouped functions can be registered and called in each part, and the atom table registration part 15 By allowing users to freely define their own variables in the atom table, general users can develop, expand, and maintain the system in the same way they would create a manual in Japanese natural language, even without programming knowledge. This makes it even easier.

FIG. 10 is a concrete explanatory diagram of the transition state of Japanese knowledge by the Japanese language understanding device, in which (A) is the output of the knowledge acquisition unit 1, (B) is the output of the knowledge compiler unit 2, and (C) is the output of the knowledge acquisition unit 1.
) is the output of the C language compiler and linker section 4. By using the above-mentioned Japanese language understanding device, for example, CAD design and drawings that were previously done manually can now be automatically done by simply changing the condition parameters by inputting know-how into the knowledge acquisition unit 1. Existing CA for CAD design and automatically outputting CAD drawings
Since the command group of the D system is output, it is possible to automate CAD and accumulate know-how on design problems as a Japanese manual.

[0026]

[Effects of the Invention] As explained above, according to the present invention, there is a knowledge compiler unit that analyzes input Japanese knowledge and converts it into a C language program, and a knowledge compiler unit that analyzes input Japanese language knowledge and converts it into a C language program. A knowledge acquisition section is provided to expand the section and acquire Japanese knowledge that is input to the knowledge compiler section, and a screen input section is provided to input atom data, which is one of the Japanese knowledge, for application execution. The configuration allows business execution programs to be created by inputting Japanese natural language, so general users who are business experts can use the system without any knowledge of programming.
In the same way as creating a manual in Japanese natural language,
It has the excellent effect of enabling system development, expansion, and maintenance.

Furthermore, when inputting object data to the knowledge acquisition section, the efficiency of lexical analysis by the lexical analysis section of the knowledge acquisition section is improved by not allowing solid writing, which is difficult to analyze in Japanese. , refer to a meaning table in which conjunctions, particles, operators, etc. are registered, extract conditions, parentheses, loop relationships, etc. in advance, and use the syntax analysis section to extract the nested structure at any time based on information from the lexical analysis section. If it is configured to improve parsing efficiency through processing, it can easily be used at the personal computer level.

[0028] Furthermore, if multiple expressions with the same meaning or user-specific expressions can be registered in the meaning table, and new functions and existing systems can be registered in the predicate table of the predicate table section, language analysis can be improved. The framework can be expanded to support user-specific applications in a general-purpose manner. In addition, it is possible to register a user's own data format in the frame table, and it is possible to register data contents in accordance with the format of the frame table from the frame data input section of the knowledge acquisition section, and the object data input section of the knowledge acquisition section. Hierarchically manage the Japanese input from , make it possible to call lower-level functions within the Japanese knowledge, register a certain set of functions and call them in each part, and add user-specific variables to the atom table. By configuring the system to be freely definable, it will be easier for general users to develop, expand, and maintain the system, just like creating a manual in Japanese natural language, even without programming knowledge. .

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a Japanese language understanding device according to an embodiment of the present invention.

FIG. 2 is a block diagram of a knowledge acquisition unit.

FIG. 3 is a block diagram of the language analysis section of the knowledge compiler section.

FIG. 4 is a block diagram of a basic library section.

FIG. 5 is an explanatory diagram of Japanese knowledge.

FIG. 6 is an explanatory diagram of object data.

FIG. 7 is an explanatory diagram of a meaning table.

FIG. 8 is an explanatory diagram of a frame table.

FIG. 9 is a flowchart of business development realized by the present invention.

FIG. 10 is a concrete explanatory diagram of the transition state of Japanese language knowledge by the Japanese language understanding device in one embodiment of the present invention.

FIG. 11 is a flowchart of conventional business development.

FIG. 12 is a flowchart of conventional system development.

[Explanation of symbols]

1 Knowledge acquisition unit 2 Knowledge compiler unit 6 Screen input unit 11 Meaning table registration unit 12 Predicate table registration unit 13 Frame table registration unit 14 Object data input unit 15 Atom table registration unit 16 Frame data input unit 17 Meaning table unit 18 Predicate table unit 19 Frame table section 21 Atom table section 31 Lexical analysis section 32 Syntax analysis section 33 Semantic analysis section 34 C standard function section

Claims (5)

[Claims] Claim 1: A knowledge compiler unit that analyzes input Japanese knowledge and converts it into a C language program; and a knowledge compiler unit that expands the knowledge compiler unit by inputting Japanese natural language and inputs the knowledge compiler unit. A knowledge acquisition unit acquires the Japanese language knowledge, and a screen input unit inputs atom data, which is one of the Japanese language knowledge, to execute the application. A Japanese language understanding device characterized in that it is configured to create a business execution program by [Claim 2] The knowledge compiler unit includes a lexical analysis unit that performs lexical analysis of input Japanese knowledge, a syntactic analysis unit that performs syntactic analysis based on the result of the analysis by this lexical analysis unit, and and a semantic analysis unit that semantically analyzes the simple sentences extracted by the knowledge acquisition unit.
a semantic table registration unit for extending the lexical analysis unit; a predicate table registration unit for extending the semantic analysis unit; and a frame table for registering a frame table that is one of the frameworks of Japanese knowledge. a registration unit; an object data input unit for inputting object data that is one of the Japanese knowledge; and an atom table registration unit for registering an atom table that is one of the frameworks of the Japanese knowledge; 2. The Japanese language understanding device according to claim 1, further comprising a frame data input unit for inputting frame data that is one of the Japanese language knowledge. 3. When inputting object data to the knowledge acquisition unit, the efficiency of lexical analysis by the lexical analysis unit is improved by not allowing solid writing, which is difficult to analyze in Japanese, and the lexical analysis unit , particles, operators, etc. are registered in the meaning table, conditions, parentheses, loop relationships, etc. are extracted in advance, and the syntactic analysis section extracts the nested structure as needed based on the information from the lexical analysis section. 3. The Japanese language understanding device according to claim 2, wherein the Japanese language understanding device is configured to improve parsing efficiency by performing the following steps. [Claim 4] Claim 2, characterized in that a plurality of expressions having the same meaning and user-specific expressions can be registered in the meaning table, and new functions and existing systems can be registered in the predicate table. Japanese language understanding device described in . 5. A user's own data format can be registered in the frame table, and data contents conforming to the format of the frame table can be registered in the frame table from a frame data input section,
In addition, it manages Japanese input from the object data input section hierarchically, allows you to call lower-level functions within Japanese knowledge, register a certain set of functions and call them in each section, and atom table 3. The Japanese language understanding device according to claim 2, wherein the Japanese language understanding device is configured such that variables unique to the user can be freely defined.