JPH0765020A - Retrieving sentence generating method - Google Patents

Abstract

PURPOSE:To continuously use a consultation menu without changing it by guiding a user in consideration of the consultations of different using frequencies and then narrowing down the real intention of the user from his ambiguous purpose description. CONSTITUTION:A purpose outline input part 202 inputs the consultation purpose of the user inputted from an input device via a user interface part 201. A forestalling question sentence generating part 203 refers to a retrieving sentence example base 207 to generate a question sentence to narrowing down retrieving sentence example. The base 207 is retrieved by the term given from the part 202 and a retrieving example sentence matched with the term is extracted. Then the user is questioned whether he takes interest in an extracted term or not. The answer of the user is received to a present, and a retrieving sentence example is narrowed down by the retrieving term with which the user designated his answer. When the decided sentence example is edited, it is additionally registered in the base 207.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and system for retrieving information by designating conditions, and is particularly useful for users who do not know the details of the information accumulated in the target system and who have difficulty in clearly indicating the retrieval conditions. The present invention relates to a particularly effective search sentence generation method when it is necessary to guide the user.

[0002]

2. Description of the Related Art With the development of computers and their peripheral technologies, various kinds of information have been accumulated in computers in large quantities. Along with this, methods for users in various fields to retrieve information in various forms have been proposed. One of them is mechanization support of consultation work, which is (1) a method of storing knowledge for classifying consultation contents and realizing it as an expert system (for example, describe knowledge of inheritance consultation as rules). Method: Nakamura, et al., "Inheritance consultation expert system," Hitachi review, Vo
l. 70, No. 11), and (2) A method of realizing as an information search system on the assumption that the content of the consultation can be expressed as a search condition (for example, considering the case where the setting of the search term is insufficient, conditions that are likely to be related) How to guess:
Japanese Unexamined Patent Publication No. 64-64032 "Information Retrieval System")
It has been known.

[0003]

The method of (1) above is
The system designer must analyze the contents of the consultation and organize knowledge in advance such as how to narrow down the problems of the consultation, such as rules, so that there are the following problems.・ Every time there is a change in the consultation menu, the work of changing knowledge will occur. -When there are consultations that are frequently used and consultations that are not frequently used, this distinction is also defined as prior knowledge, so it cannot be adapted according to the operational status of the system. On the other hand, the method (2) above is incomplete for the user,
Since it is premised that search conditions can be explicitly described,
There are the following problems. -The true intention cannot be narrowed down from the vague description of the user's purpose.・ It is not possible to narrow down the intention from matters that are completely unrelated to input. It is an object of the present invention (1) to provide a method that enables continuous use even if the consultation menu is changed without changing knowledge. (2) It is to provide a method for guiding a user in consideration of frequently used consultation and less frequently used consultation. (3) It is to provide a method capable of narrowing down the true intention from the vague description of the purpose of the user.

[0004]

In order to achieve the above object, the present invention provides a search statement case base storing a plurality of search statement cases and a search term attached to each search statement case and extracted from the search statement case. In a search system including a processing device, a main storage device, a display device, and an input device, a search term is extracted from a purpose summary input by a user, and the search sentence case base is extracted based on the extracted search term. Is generated, and a question sentence for narrowing down the search target is generated and displayed based on the search term attached to the retrieved retrieval sentence case and the retrieved retrieval sentence case, and the retrieval is performed based on the user's response to the question sentence. The search sentence cases are narrowed down, the narrowed down search sentence cases are displayed for the user to select, and the search sentence case selected by the user is set as the search sentence. Furthermore, when the number of search sentence cases as a result of the narrowing down is not narrowed down to a predetermined number or less, generation and display of the question sentence and narrowing down of search sentence cases are performed.
The search is repeatedly performed until the number of search sentence cases as a result of the narrowing down falls below a predetermined number. Further, when a user selects a near miss search sentence example, the near miss search sentence example can be edited on the display screen, and the search sentence example edited by the user is used as a search sentence, and the search term is searched from the search sentence. I am trying to extract. Further, the search sentence obtained by the editing and the search term extracted from the search sentence are additionally registered in the search sentence case base. Further, the usage frequency of the search statement case is stored in the search statement case base in correspondence with the search statement case, and the use frequency corresponding to the search statement case is updated every time the search statement case is selected. .

[0005]

In the present invention, (1) in the step of inputting the purpose outline, the vague purpose of the user is accepted, not the retrieval statement executed immediately. (2) In the step of generating a question sentence for narrowing down the search target from the input purpose outline, what question should be asked to narrow down the true intention of the user is processed. (3) In the step of displaying the search sentence example of the user's search intention, which is estimated from the above-mentioned outline of the purpose and the response, the user can confirm whether or not the user has the true intention regarding his / her consultation. (4) In the step where the user edits the search sentence case, if the case is different from the case, the case can be corrected to express the case. With these, the true meaning of the consultation can be narrowed down from the vague consultation purpose. Furthermore, by (5) accumulating the edited search sentence as a new search sentence example, new knowledge can be added sequentially. (6) When displaying the search sentence, Determine the display order of search sentences based on the frequency of use. As a result, it is possible to follow the interests of the user, and according to this simple method, there is no need to describe and maintain knowledge about the consultation method in the rules.

[0006]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail below with reference to the drawings. FIG. 2 is an example of a functional block diagram of a search system for implementing the present invention. The system includes two storage units, a search sentence case base 207 and a database 209. As shown in FIG. 5, the search sentence case base is composed of several cases (here, 510, 520, ..., 550, ...). One case 510 is case number 501, search sentence 5
02, a search term 503 extracted from the search sentence, and a frequency 504 at which the case is used. Database 209
May be a document database that is the target of full-text search or a relational database. The processing unit includes a user interface unit 201, a purpose outline input unit 202, a preceding question generation unit 203, and a search sentence example extraction unit 2
04, search sentence example selection / editing unit 205, search sentence execution unit 2
06, and a search sentence example updating unit 208.

FIG. 9 shows a hardware configuration of a search system to which the present invention is applied. Each workstation (W
S) 901 includes a processing device 910, a main storage device 911, a display device 912, and an input device 913. Server 902
Database 209 and search sentence case base 207
Is placed. The processing in the processing unit of the functional block diagram of FIG. 2, which will be described in detail later, is performed by the processing device 910 and the main storage device 911 of each WS, and the display to the user is performed by the display device 912. Is input by the input device 913. Although all of the above processing may be performed by the processing device 910 and the main storage device 911, the database 209 and the search statement case base 207 are also included.
It is desirable to place the management system for the server on the server 902. Further, when the database 209 and the search statement case base 207 are large in number and distributed, each of the one or more computers 904 may be connected as a server 905. In this case, the workstation (WS) is connected to the computer 904 via the gateway 902 and the switching network 906.

FIG. 1 is a flow chart showing an embodiment of the present invention. First, in step 101, the purpose outline input unit 202 inputs the consultation purpose from the user, which is input from the input device, via the user interface unit 201. In the embodiment, it is assumed that the user cannot fully express his or her wish at this point and can only enter an outline of the purpose. Conversely, if only the outline is input, the true meaning is estimated and a search sentence is generated by the following processing. Input for this purpose may be made by selecting from pre-made menus, by entering words such as keywords, or by entering a natural language such as "I want to hire an elderly person". It does not matter which method is used. In the last case, the term "elderly person" in which the user is interested is extracted by the purpose outline input unit by morphological analysis known in natural language processing technology. Hereinafter, in the present embodiment, the purpose of consultation from the user is represented by one or more terms. next,
In step 102, the advanced question sentence generation unit 203 refers to the search sentence case base 207 to generate a question sentence for narrowing down the search sentence cases. Details of this processing are shown in FIG.

In FIG. 3, in step 301, the search sentence case base 207 is searched by the term (for example, "elderly person") from the purpose outline input unit 202, and the search case sentence that matches the term is extracted. For example, the case 51 shown in FIG.
0 to case 550 all meet this condition. In step 302, the frequency of appearance of each term in the search sentence examples is calculated for each search term 503 associated with each retrieved search sentence. FIG. 6 shows an example of the tabulation status. Here, the term “employment” 601 is “24 cases” 602 in the number of cases “254 cases” 603 searched by the term “elderly person”.
It indicates that the frequency is. In step 303,
Extract terms with frequencies close to 50%. Here, 50% means that the search cases are the most balanced, regardless of whether the user has “related” or “not related” to the term.
Because it is divided. In step 304, whether the user is interested in the extracted term is displayed as a diploma on the extracted term via the user interface unit 201 to ask the user a question. Note that more than one term may be used for a question. Also, how many questions to ask may be determined from the number of search sentence cases at that time. The number of terms to be queried can be determined using an information theory coding algorithm. FIG. 10 shows a display example of the question displayed on the display screen.

Returning again to the flow of FIG. 1, step 103
Then, the user's response to the above question is accepted, and the search sentence cases are narrowed down by the search term designated by the user as a response. Next, in step 104, a predetermined threshold value (for example, 10) is referred to with reference to the narrowed-down number of cases.
If it is larger, the process returns to step 102 and the narrowing is repeated. If it is less than or equal to the threshold value, the process proceeds to step 105.
In step 105, the cases narrowed down by the search sentence case extracting unit 204 are sorted by the use frequency 504, the search sentence 502 is displayed on the screen via the user interface unit 201, and if there is an appropriate case for the user. If there is a close case that is not suitable (near miss), ask the close case (that is, a case that can be edited to show the true meaning of the user) and apply. If there is no case, ask them to do so.
FIG. 7 is a display example of this search sentence example. In the embodiment, this search sentence example is treated as "what the user really means".

At step 106, the response of the user to the displayed case is determined. As a result of the determination, when the response is that "the appropriate case has not been displayed", the process returns to step 101 for inputting the consultation purpose, and the process is repeated from step 101. When there is an appropriate case that represents the true meaning and a response is made to instruct the case, in step 109, the usage frequency 504 of the instructed search statement case is incremented by 1, and then the search statement case update section 208 is added. Then, the search sentence case base 207 is updated and the process proceeds to step 110. If there is a case corresponding to a near miss and the case is designated, the search sentence case selection / editing unit 205 edits the case in step 107. That is, by selecting the instructed case via the user interface unit 201, for example, the user can edit the image on the screen using the input device, and the user edits it.
The edited result is input via the user interface unit 201. When the editing is completed, the search statement case example updating unit 208 additionally registers in the search statement case base 207 in step 108. This processing will be described with reference to FIG. If there is a near-miss case in step 106 of FIG. 1, the case is edited in step 107. The edited case can be reused by adding it to the search sentence case base. .

In step 401, the edited case example of the search sentence is morphologically analyzed and word-divided. In step 402, some predefined words of the divided words are removed as unnecessary words (for example, "want to see" and "want to know"). The word after being removed becomes the search term 503 of the edited search sentence example. And step 40
In 3, the usage frequency 504 of the edited search sentence example is set to 1. Finally, in step 404, the search sentence example update unit 2
08, the edited search sentence case, its search term, and frequency of use are additionally registered in the search sentence case base. In other words, in the search sentence case base, the edited search sentence case is additionally registered,
In addition, the frequency of use of the reused search sentence examples is updated, so that the qualitative improvement is successively achieved.

Returning again to the flow of FIG. 1, step 110
Then, the search statement execution unit 206 searches for corresponding information from the database 209 according to the search term of the selected or edited search statement case, and the user interface unit 20
Information is provided to the user via 1. In this embodiment, the information is searched based on the search terms shown in FIG. 5, but at this time, either the logical sum or the logical product of the terms may be searched, or the information including many terms may be searched. Good. For terms, it is desirable to eliminate the omission of search by using the synonym table of FIG. 8 as is well known.

[0014]

As described above, according to the present invention, even when there is a change in the consultation menu (search sentence case-based data), it can be continuously used without changing knowledge. It is possible to obtain the remarkable effect that it is possible to guide the user in consideration of frequently used consultation and rarely used consultation, and to narrow down the intention from the vague description of the user's purpose.

[Brief description of drawings]

FIG. 1 is a diagram showing a flowchart showing an embodiment of the present invention.

FIG. 2 is a functional block diagram of a search system that implements the present invention.

FIG. 3 is a diagram showing a flowchart of processing for generating a question text for narrowing down search text examples.

FIG. 4 is a diagram showing a flowchart of a process of additionally registering an edited search statement case in a search statement case base.

FIG. 5 is a diagram showing an example of a search sentence example. ,

FIG. 6 is a diagram showing an interim aggregation status of frequency by term of each term appearing in a search sentence example.

FIG. 7 is a diagram showing a display example of a search sentence example.

FIG. 8 is a diagram showing an example of a synonym table.

FIG. 9 is a diagram showing a hardware configuration of a search system to which the present invention is applied.

FIG. 10 is a diagram showing a display example of a question displayed on a display screen.

[Explanation of symbols]

 901 workstation 902, 905 server 903 gateway 904 computer 906 switching network 910 processing device 911 main storage device 912 display device 913 input device

Claims (5)

[Claims] 1. A search statement case base storing a plurality of search statement cases and a search term associated with each search statement case and extracted from the search statement case, a processing device, a main storage device, and a display device, A search statement generation method in a search system including an input device, wherein a search term is extracted from a purpose summary input by a user, the search statement case base is searched by the extracted search term, and the searched search statement is searched. A question sentence for narrowing down the search target is generated and displayed based on the search term attached to the case and the searched sentence case, and the retrieved search case case is narrowed down based on the user's response to the question sentence. A search statement generation method characterized by displaying the narrowed-down search statement case so that the user can select it, and using the search statement case selected by the user as a search statement. 2. The search sentence generation method according to claim 1, wherein when the number of search sentence cases as a result of the narrowing down is not narrowed down to a predetermined number or less, generation and display of the question sentence and search sentence cases are performed. A search sentence generation method characterized in that the narrowing down is repeated until the number of search sentence cases as a result of the narrowing down becomes a predetermined number or less. 3. The search sentence generation method according to claim 1, wherein when a user selects a near miss search sentence case, the near miss search sentence case can be edited on the display screen, and the user can edit the near miss search sentence case. A search sentence generation method, wherein a search sentence example edited by is used as a search sentence, and a search term is extracted from the search sentence. 4. The search statement generation method according to claim 3, wherein the search statement obtained by the editing and the search term extracted from the search statement are additionally registered in the search statement case base. Search sentence generation method. 5. The search statement generation method according to claim 1, wherein a usage frequency of the search statement case is stored in the search statement case base corresponding to the search statement case, and a search is performed. A method for generating a search statement, characterized in that the frequency of use corresponding to the search case is updated every time a case is selected.