JPS63219034A - Natural language interaction device - Google Patents

Abstract

PURPOSE:To operate an objective region through an interaction in a natural language, without the knowledge of a technical term by providing a means, which stores the correspondence relation between the natural language and a meaning expression to express a technical knowledge, and obtains the technical knowledge from an input sentence. CONSTITUTION:When a user inputs a question through the natural language by an input device 1, a syntax.meaning analyzing device 2 generates the meaning expression of it. A topic selecting device 4 searches a rule, embedded in the meaning expression, among contents, stored in a topic.condition correspondence table storing device 5, and makes a not fulfilled part be the candidate of the meaning expression of the contents, to inquire of the user next time. A sentence generating device 8 outputs a Japanese sentence, corresponding to the meaning expression of the contents to inquire of the user next time, and displays it by a display device 9. When the applicable rule does not exist any longer, and there is no more candidate, the topic selecting device 4 searches a data base 7, and delivers the resulted meaning expression to the device 8, and displays the Japanese sentence, corresponding to a solution, on the display device 9.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a natural language dialogue device for performing question answering, searching, and the like.

BACKGROUND OF THE INVENTION Conventional natural language dialogue devices for question answering, searching, etc. generate sentences using specialized terminology specific to the problem area to be answered, searching, etc. They used a device that interpreted the options and specialized sentences.

A commonly used database search device using natural language will be described.

Database searches in Japanese are usually performed in the form of logical sums and logical products of keywords. Experiments are being carried out to extract keywords from natural language, perform database searches based on them, and display or generate answers. For example, from the phrase "Technology introduction contract where the licensor is Company A" (1), the logical expression of the keyword licensor (Company A) * Technology introduction contract (11) is written as f(r noun 1? is noun 2? ") → Noun 1? (Noun 2?) f (Dependent sentence? - Noun?) → f (Dependent sentence?) * f (Noun?) However,? Words marked with a mark represent character strings that can be classified into that syntactic category, and ° represents a concatenation of character strings.

It is generated using a syntax/logical expression correspondence rule f defined as follows. Such conversion is effective only when words corresponding to attributes that are already defined in the domain of the database are used directly.

Problems to be Solved by the Invention Conventional natural language dialogue devices generate sentences using technical terminology specific to the problem area used for question-answer searches, etc. This method has the problem that it is difficult to use for users who do not have specialized knowledge of the subject area.However, the present invention can be operated even by users without specialized knowledge. The purpose is to make it possible.

Means for Solving the Problems The natural language dialogue device of the present invention has a means for obtaining a semantic expression of a user's natural language input sentence, and a correspondence relationship between the semantic expression and a semantic expression representing specialized knowledge. a means for storing; a means for obtaining specialized knowledge from the semantic representation of the input sentence and the correspondence; and a means for obtaining specialized knowledge from the semantic representation of the input sentence and the correspondence; By having a means to obtain natural language and a means to generate natural language from the finally obtained semantic expression, it is possible for users without specialized knowledge of the subject area to perform question answering, searching, etc. through natural language dialogue. shall be.

Function: Using the above-mentioned methods, it analyzes the natural language input by the user, derives specialized knowledge corresponding to the user's intention and situation,
At the same time, it generates sentences related to the intention and situation to ask the user next.

EMBODIMENT OF THE INVENTION The present invention will be described in detail with reference to FIG. 1 regarding the case where a database search regarding contract documents is executed through dialogue in Japanese.

The embodiment includes an input device 1 such as a keyboard for inputting Japanese, and a syntactic/semantic analyzer 2 used to perform syntactic/semantic analysis of input Japanese to obtain its semantic expression.
, a topic/condition correspondence table storage device 5 that stores topic/condition correspondence rules representing the correspondence between a series of semantic expressions based on concepts provided by the user and a specialized concept regarding the contract (based on the semantic expressions); The meaning obtained by the condition extraction device 3 and the syntactic/semantic analyzer 2 is used to obtain the semantic expression that becomes the condition for database search from the semantic expression obtained by the syntactic/semantic analyzer 2 and the above-mentioned topic/condition correspondence rules. A topic selection device 4 for obtaining a semantic expression of the next question to be asked to the user from the expression and the above-mentioned topic/condition correspondence rules, a database 7 related to contract documents, and database search conditions obtained by the condition extraction device 3. A database search device 6 generates a command for database search from the semantic expression, searches the database 7, and obtains the semantic expression of the search result.The topic selection device 4 then asks the user the following questions. a sentence generation device 8 for generating, in Japanese, a semantic expression of the content to be performed and a semantic expression of the database search results obtained by the database search device 6; a VDT for displaying the Japanese language obtained by the sentence generation device 8; It is composed of a display device 9 such as.

Below, in the explanation, the part surrounded by single quotation marks (゛) indicates a semantic expression description.

In the topic/condition correspondence table storage device 5 of the device of this embodiment,
Topic/condition correspondence rules in the format of "left side → right side" are stored in advance. The left side of the rule is written as a sequence of semantic expressions with variables based on concepts provided by the user, and the right side is written based on specialized concepts regarding contracts (if the semantic expressions have the same variables as the left side). Below, in the rules, a question mark (?) is placed after a capital letter.
) indicates variables. Current topics/
In the condition correspondence table storage device 5, 'Sell (subject: X?, target: technology) → 2 licensor (X?)''Use (subject: ''Sue (Subject: X72 Object: Y?) → 'Licensor (X?)'.

'Licensee (Y?)' A large number of rules are stored.

FIG. 1 shows that at this time, the user inputs "A" using the input device 1.
``I want to know the contract under which the company sold its technology'' is input, and it is output as a character string 11. The syntactic/semantic analyzer 2 receives the above character string 11. , I want to know the meaning of the user of the sentence (21)' (Object: Contract ( Content: Sell ( Subject: Company A.

Target: Technology) ).

and syntactic theme 22 techniques are output as shown in Figure 2.

The condition extracting device 3, as shown in FIG.
, the left side of the rule or the semantic expression 21 among the topic/condition correspondence rules stored in the topic/condition correspondence table storage device 5
The right-hand sides of all such rules are collected to generate a semantic expression 31 that is a condition for database search. In this case, if there are no other applicable rules, the semantic expression 31 becomes 'Licensor (Company A)'. The semantic expression 31 is stored in the database search device 6 as a condition for database search.

At the same time, as shown in FIG. 4, the topic selection device 4 uses the above-mentioned semantic expression 21 to select 'sell (subject: X?

Target: Technology (Field: Y?)'' ``Licensed product (Y?)'' Search for rules where only a portion of the left side of the rule is embedded in the semantic expression 21, and find the unsatisfied portion of the left side. Technology (Field: Y?)' as a candidate for the semantic expression of the content to be asked to the user next. If there is one 0 candidate, use it as the semantic expression 41 of the content to be asked next to the user.・If there are multiple candidates, use the syntactic theme 2 above.
A candidate having the same syntactic theme as 2 is selected and used as the semantic expression 41 of the next question to be asked to the user. In this case, the syntactic subject 22 is "technology" and the syntactic subject of 'technology (field: Y?)' is also "technology", so the semantic expression 41 is 'technology (field: Y?)'. becomes.

The sentence generating device 8 in FIG.
``What is the technology related to?'' is output and displayed to the user through the display device 9.

In the situation shown in FIG. 4, if there are no applicable rules and no candidates, the topic selection device 4
An instruction 42 instructing the start of database search is output. Then, as shown in FIG. 6, the database search device 6 converts the stored search condition into a database search instruction 61, searches the database 7, and converts the result into a sentence as a semantic expression 62 of the database search result. It is passed to the generation device 8. Below, the same processing as in FIG. 5 is performed, and the Japanese sentence corresponding to the answer is displayed on the display device 9.

In FIG. 4, when the topic selection device 4 selects one topic from a plurality of candidates, the database search device 6 preferentially uses a rule that has a particularly required condition on the right side. It is also possible to use the techniques simultaneously.

Effects of the invention It is possible for the user to perform operations related to the target area through dialogue using natural language, as well as knowledge of technical terminology related to the target area.

[Brief explanation of drawings]

1 to 6 are block diagrams of a database search system according to an embodiment of the present invention. 1...Input device, 2...Syntax/semantic analysis device, 3...Condition extraction device, 4...
Topic selection device, 5... topic/condition correspondence table storage device, 6... database search device, 7...
...Database, 8...Sentence generation device, 9...
...Display device. Applicant Director of the Agency of Industrial Science and Technology Yuki Iizuka Santsutsu Figure 1 Figure 2 Figure 3 Figure 4 Figure 5

Claims (1)

[Claims] means for obtaining a semantic representation of a user's natural language input sentence; means for storing a correspondence relationship between the semantic expression and a semantic expression representing specialized knowledge; A means for obtaining specialized knowledge from the relationship, a means for obtaining a semantic expression for the next question to the user from the semantic expression of the input sentence and the correspondence relationship, and a means for generating natural language from the finally obtained semantic expression. A natural language dialogue device having means for generating.