JPH02189680A - Information retrieving system - Google Patents

Abstract

PURPOSE:To reduce the probability of retrieval failure by selecting keywords in stages, concentrating retrieved results, and when the number of retrieved results is zero, advancing retrieval by utilizing an upper concept and relaxing a condition. CONSTITUTION:The retrieving system consists of an inquiry input part 2 for receiving an inquiry 1 from a user and converting the inquiry into a retrieving condition, a keyword selection part 4 for selecting keywords in stages while deciding retrieved results, a retrieval formula forming part 5 for forming a retrieval formula by using the selected keyword, a retrieval execution part 6 for executing data base retrieval by the formed retrieval formula, and a response forming part 10 for forming a response. The retrieved result is fed back to the keyword selection part 4, the retrieving conditions are successively added to the retrieved result or retry is executed by using the upper concept to execute retrieval in stages. Consequently, retry when the number of retrieved results is zero can be easily executed and the probability of retrieval failure can be suppressed to the minimum.

Description

[Detailed Description of the Invention] (Industrial Application Field) The present invention performs a search in stages by checking the search results, adding search conditions one by one, or retrying using a superordinate concept. This relates to the information retrieval method used.

(Prior art) In database searches using commonly used keywords, keywords to be used in the search formula must be selected using a thesaurus, etc., and a keyword logical formula must be created, making it difficult for inexperienced users to use. It is.

In response, a method has been proposed in which a keyword logical formula is automatically generated from a user's inquiry input and a search is executed. For example, the technology described in Information Processing Society Natural Language Processing Study Group Material 58-8 (1986), pp. 1-8, [Information Retrieval System Based on Natural Language Understanding IRIS J] is known. This is from a natural language sentence,
A content model representing the semantic content is created, and logical expressions are generated from keywords from the content model. Figure 2 shows an example of a content model. (a) represents a question, (b) represents a content model corresponding to (a), and (e) represents a keyword logical formula generated from (b).

(Problem to be solved by the invention) However, in the conventional method, the content model shown in Figure 2 (b) is converted into a single keyword logical formula, and a search is performed once using that logical formula. finish. At this time,
1) In order to avoid search failures, keywords are also OR-combined with higher-level concepts, so the searched data will contain many redundant items. 2) Search failures with such keyword logical formulas There is a problem in that if you do so, the program will end immediately.

An object of the present invention is to provide an information search method that prevents such an increase in search results and minimizes the probability of search failure.

(Means for solving the problem) An inquiry input unit that accepts inquiries from users and converts them into search conditions, a keyword selection unit that selects keywords step by step while determining search results, and a keyword selection unit that uses the selected keywords. It consists of a search expression generation unit that generates a search expression, a search execution unit that executes a database search using the generated search expression, and a response generation unit that generates a response.
The search results are fed back to the keyword selection section, and the search is performed step by step by adding search conditions one by one or retrying using superordinate concepts.

(Operation) According to the present invention, the search is performed step by step by sequentially adding search conditions while checking the search results and retrying using superordinate concepts. This makes it easier to redo the case.

(Example) FIG. 1 is a block diagram showing one embodiment of the present invention.

Inquiry 1 is an inquiry manually made by the user.

For example, [What kind of laptop computer is released by Company A? ”
Possible questions include: Examples of human-powered methods include human-powered input using natural language text and input using menu selection.

The inquiry input unit 2 is a general-purpose interface that receives the inquiry 1 and generates the search condition 3. Third
The figure schematically represents an example of search conditions generated from the above-mentioned question example.

The keyword selection section 4 refers to the search conditions 3 and the thesaurus 7, selects keywords, and passes the selected keywords to the search formula generation section 5.

FIG. 4 shows the processing flow of the keyword selection section 4. First, in step 41, the search history 12 as shown in FIG. 5 is referred to to determine whether the search was successful or unsuccessful. FIG. 5(a) shows the initial state of the search history, and FIG. 5(b) shows the intermediate state of the search history.

Pointer 51 is a pointer that holds the search number executed immediately before, and pointer 52 is a pointer that holds the search number that was successfully executed immediately before. In (a), the search number O is set for both the pointer 51 and the pointer 52, and the maximum number that can be given, for example, '9999' is set for the number of search results. If the number of search results corresponding to the number is 0, it is determined that the search has failed, and if it is not 0, it is determined that the search has been successful.If the search fails, at 42, refer to the thesaurus 7 as shown in FIG. It is determined whether a superordinate concept of the selected keyword exists. If it exists, the superordinate concept is fed back to the search formula generation unit 5 as the next keyword in step 43. If there is no superordinate concept, the keyword selection unit , and issues a search failure response request to the response generation unit 10. If the search is successful in the check at 41, it is checked at 45 whether there are any unused condition items in the search conditions 3. If all the conditions are satisfied, a response request indicating a successful search is issued.If there is a condition item that has not been used yet, the condition item is specified in step 46, and the condition value of that item is used as the next keyword.

The search expression generation unit 5 extracts synonyms of the keyword passed from the keyword selection unit 4 from the synonym dictionary 8, and generates a logical expression by combining each synonym with OR. Furthermore, a logical expression is generated by ANDing this logical expression and the keyword logical expression corresponding to the search number indicated by the pointer 52 in the search history, and is passed to the search execution unit 6 as a new search expression. If the pointer 52 indicates search number 0, the keyword logical formula passed from the keyword selection unit 4 is passed as is to the search execution unit 6.

The search execution unit 6 executes a search on the database 9 using the search formula passed from the search formula generation unit 5, stores the search number, the number of searches, and the keyword logical formula in the search history 12, A next keyword selection request is sent to the selection unit 4.

Database 9 is a database to be searched.

The response generation unit 10 generates a response 11 in accordance with the response request from the keyword selection unit 4. The response requests include a search success response request and a search failure response request. FIG. 7 shows an example of a response to the above-mentioned example question.

FIG. 7(a) shows an example of a response to a response request for a failed search, and FIG. 7(b) shows an example of a response when the search is successful and there is one search result.
C) is an example of a response to a response request when the search is successful and there are multiple search results. Possible output formats include natural language sentences, tables, and graphs.

(Effects of the Invention) As described above, according to the present invention, search results are narrowed down by selecting keywords in stages, and if there are no results, the search results are relaxed by using higher-level concepts. Since the search progresses, the probability of a search failure is reduced, and the search results do not contain too much redundant data due to too loose conditions.

Furthermore, since the search history is saved, it becomes easy to know under what conditions the search failed.

[Brief explanation of the drawing]

FIG. 1 is a block diagram representing an embodiment of the information retrieval method of the present invention. 1: Inquiry, 2: Inquiry input section, 3: Search condition, 4: Keyword selection section, 5: Search expression generation section, 6: Search execution section, 7: Thesaurus, 8: Synonym dictionary, 9: Database, 10: Response generation unit, 11: response, 12: search history. FIG. 2 is a diagram showing an example of a content model. 3 is a diagram showing an example of search conditions, FIG. 4 is a diagram showing the processing flow of the keyword selection section, FIG. 5 is a diagram showing an example of the search history, FIG. 6 is a diagram showing an example of the thesaurus, FIG. 7 is a diagram showing an example of a response.

Claims (2)

[Claims] (1) An inquiry input section that accepts inquiries from users and converts them into search conditions, a keyword selection section that judges search results and selects additional keywords based on the judgment results, and a search formula using the selected keywords. It consists of a search expression generation section that generates a search expression, a search execution section that executes a database search using the generated search expression, and a response generation section that generates a response, feeds back the search results to the keyword selection section, and refers to the search results. An information retrieval method characterized by performing a step-by-step search. (2) An inquiry input section that accepts inquiries from users and converts them into search conditions, a keyword selection section that judges search results and selects keywords of superordinate concepts if the judgment results are empty, and selected keywords. It consists of a search expression generation section that generates a search expression using the search expression, a search execution section that executes a database search using the generated search expression, and a response generation section that generates a response, and feeds back the search results to the keyword selection section. An information retrieval method characterized in that a search is performed step by step with reference to the search results.