JPH11232297A - Information retrieving method, information retrieving device and storage medium in which information retrieving program is stored - Google Patents

Abstract

PROBLEM TO BE SOLVED: To efficiently decide a proper field especially when an object of user's interest is extended to plural fields and when user's intention related to fields with wide range is inputted in information retrieval which is categorized into the fields. SOLUTION: A fact and a question are generated about an inputted event and set in a work memory 28 by using event knowledge 31 formed by collecting knowledge about the event corresponding to the user's intention and a set 32 of rules formed by collecting rules to be expressed by a conditional part and an action part. When the rule whose conditional part is established by the generated fact exist in the set 32 of rules, the rule is extracted and executed. The fact obtained by execution is set in the work memory 28. In addition, a user is inquired about the generated question again, input is prompted, the fact is generated from an input result and set in the work memory 28. When no question and rule exist in the work memory 28, the fact to express the field to be served is extracted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for retrieving information categorized into fields, and particularly to a case where an object of interest covers a plurality of fields or a user's intention related to a wide range of fields is input. More particularly, the present invention relates to an information search method, an information search device, and a storage medium storing an information search program for efficiently determining an appropriate field.

[0002]

2. Description of the Related Art An information search method using a natural language is a method by which an unspecified number of users can easily obtain information on a target field (hereinafter referred to as a service) in a yellow page service. For this method, there is a field inference method in which a keyword of "product + verb" is extracted by performing morphological analysis, semantic analysis, and intent understanding from a user's input in a natural language, and mapping the keyword to a service.

FIG. 5 shows a configuration diagram of a service mapping unit in a conventional field inference system. The field name is input from the input unit 11 to the service mapping unit 12. The service mapping unit 12 includes a field / service correspondence table 14 and a field / service matching means 13, and the field / service matching means 13 uses the field / service correspondence table 14.
The service corresponding to the field is derived and output to the output unit 15. In this way, services corresponding to the field can be derived.

In this field inference method, for example, a one-interaction method of inferring "Izakaya" in response to an input of "want to drink" and guiding the result directly to the user has been adopted.

[0005]

As described above, when mapping an inferred field to a service, the service mapping unit 12 in the conventional field inference method uses the field / service correspondence table 14 to provide a service. Is fixedly decided. As a result, the user's intention (for example, in the case where the object of interest spans multiple fields, or where the guidance range is different depending on the viewpoint / case, such as in the case of travel, where accommodation is required only when the travel spans multiple days). (Hereinafter referred to as events.)

[0006] The present invention has been made in view of the above points, and an object of the present invention is to provide a service mapping unit that uses event knowledge and a rule set for a field and information acquisition by conversation with a user. An object of the present invention is to provide a means that can guide information that can flexibly cope with the above situation by performing inference and linking to a service.

[0007]

In order to achieve the above object, the present invention provides an event knowledge which collects knowledge about an event corresponding to a user intention, and a rule set which collects rules expressed by a condition part and an action part. Using the rules, a fact and a question are generated, set in the work memory, and if there is a rule in the rule set such that the fact in the work memory satisfies the condition part of the rule, the rule is extracted. If the rule is set in the memory and the rule exists in the work memory, the rule is executed and the obtained fact is set in the work memory. If the question exists in the work memory, the user is asked a question and prompted for input. The gist is to provide means for generating facts from input results, setting them in the work memory, and extracting the facts representing the service field when there are no more questions and rules in the work memory. .

[0008] A program for realizing the above processing means by a computer can be stored in an appropriate storage medium such as a computer-readable portable medium memory, a semiconductor memory, and a hard disk.

When the field inferred by the above means is an event, by using the event knowledge in the event expanding means, the known information is made into fact and the fact including unknown information is asked about the information about the event. And store it in the work memory. A rule applicable to the fact set in the work memory in the rule set is extracted by an application rule extracting unit, and a new fact is obtained by applying the rule in the rule execution unit and stored in the work memory. If there is no applicable rule for the facts in the work memory, a question is searched, and if there is a question, a question is returned to the user by the conversation guiding means, and the new fact is stored in the work memory. When there are no more applicable rules or questions, the service extraction means extracts the facts representing the service field from the work memory, so that the user's intention, that is, the guidance service differs from viewpoint to viewpoint, that is, for events, Missing information can be supplemented by conversation, and rules can lead to more reliable services.

[0010]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration of an information search device according to one embodiment of the present invention.

In FIG. 1, reference numeral 20 denotes an information search unit implemented by a CPU, a memory, and an information search program. Reference numeral 21 denotes an input unit for inputting an event or a response to a question to the information search unit 20, such as a keyboard or an input file. Reference numeral 22 denotes an output unit for displaying or outputting a service from the information search unit 20 as a question or a search result, such as a display or an output file.

The knowledge database unit 30 includes an event knowledge 31 and a rule set 32.
Performs an information search using the knowledge database unit 30. The event expansion means 23 in the information search unit 20
Using the event knowledge 31, information corresponding to the event input from the input unit 21 is output to the work memory 28. The application rule extraction unit 24 compares the rule set 32 with information in the work memory 28 and outputs a suitable rule to the work memory 28. The rule executing means 25 applies the rules in the work memory 28 to the information in the work memory 28 and outputs the result to the work memory 28. If there is a question in the work memory 28, the conversation guiding means 26 executes the question and stores the result in the work memory 2.
8 is output. The service extracting unit 27 extracts information about the service from the information in the work memory 28 and outputs the information to the output unit 22.

FIG. 2 is a diagram showing an example of event knowledge. The event knowledge is composed of two types, one event record indicating each event, and a comment record that starts with # and does not have special rules. The event record defines an event name, a parent record name of the event, an attribute name of each event, and an attribute value for the attribute. The first element is an attribute value having an attribute name of “event name”, and the second element is an attribute value having an attribute name of “upper event”. If the parent record does not exist or if only the attribute name is determined and the corresponding attribute value is not determined, fill in with a null code. The third and subsequent elements are the attribute name and attribute value of the attribute of the event.

For example, an event record 2 shown in FIG.
Event knowledge of 2 is that the event name is “Go around the shrine” and the upper event (parent event) is “Travel”.
In addition, the attribute name of the attribute of the event “visiting a shrine” is “interesting object”, which means that the attribute value is “shrine”.

FIG. 3 is a diagram showing an example of a rule set. The rule set is composed of two types, one rule indicating each rule, and a comment sentence that starts with # and has no special rules. A rule describes an attribute name and its corresponding attribute value. For each rule, a condition part indicating the rule application condition, and an action part consisting of the attribute name generated as a result of applying the rule and the corresponding attribute value are included. Describe.

For example, the rule 3-1 in FIG. 3 is that if there is a fact that the attribute value of the attribute name “place” is “Kyoto”, the attribute name is “interesting” and the attribute value is “Kyoto” Which produces the fact that

Next, an information retrieval method according to the present invention will be described with reference to the flowchart shown in FIG. In addition,
A specific example will be described using an example in which the event “shrine tour” is input from the input unit 21.

[Event Expansion Step S1] First, in the event expansion step S1, the event knowledge 31 is searched by using the event input from the input unit 21 as a key, and an event record having the same event name is extracted. In the event record, if the attribute value contains a null code, the information is not determined, otherwise the information is determined to be determined, and in fact, it is not determined from the determined information A question is generated from the information and stored in the work memory 28.

As the event knowledge 31, a parent event of each event is defined, and when a certain event is defined, the characteristics of the parent event are inherited and defined, thereby facilitating the management and definition of the event knowledge. Can be improved. In this case, the parent event name of the event record corresponding to the input event is extracted, and facts and questions are similarly generated for the event record corresponding to the parent event name and stored in the work memory 28.

By repeating such an operation until the parent event name becomes a null code, facts and questions relating to the input event and all the parent events can be set in the work memory 28.

For example, when the event "tour to shrines" is input, the event record 2-2 shown in FIG. 2 is selected, and the following facts are stored in the work memory 28. [Fact 1]: Event name (tour to shrines) [Fact 2]: Interest object (shrines) Also, since the parent event name is “travel”, event record 2-1 is selected next, and the following facts and The question is additionally stored in the work memory 28.

[Fact 3]: Event name (travel) [Question 1]: Period () [Question 2]: Location () Here, the period () and the location () are attribute values for the attribute name (). This is a question because no parameters have been set.

The case where an event is input from the input unit 21 has been described above.
Alternatively, the same processing is performed when returning from the conversation guidance process S4. That is, when a new event is generated as a result of executing a rule or input based on conversation guidance, the event knowledge 31 is searched as described above, and a new fact or question is generated. On the other hand, if a new event is not generated, the process proceeds to the application rule extraction process S2 without doing anything in this event expansion process S1.

[Applied Rule Extracting Step S2] Next, in the applied rule extracting step S2, the rule set 32 is searched, and a rule exists such that the fact in the work memory 28 satisfies the condition part of the rule in the rule set 32. If so, all the rules are stored in the work memory 28 as applicable rules.

For example, in the case of the rule set shown in FIG. 3, the rule 3-2 having the fact “interest (shrine)” in the condition part is extracted and stored in the work memory 28. [Rule 1]: if Interest (shrine) then service (shrine guide); If applicable rules are stored in the work memory 28, the rule execution process S3 is performed. If there are no applicable rules and there are questions, the conversation guidance process S
Perform 4. If neither applicable rule nor question exists, a service extraction step S5 is performed.

[Rule Executing Step S3] In the rule executing step S3, the applicable rule in the work memory 28 is executed, and the fact described in the action part of the rule is stored in the work memory 28.

For example, if the above [Rule 1] exists in the work memory 28 as an applicable rule, the following facts are added to the work memory 28. [Fact 4]: Service (Shrine Information) Here, [Rule 1] in the executed work memory 28 is
Since it does not need to be executed again,
Distinguish from unexecuted rules.

[Conversation Inducing Step S4] In the conversation inducing step S4, a question in the work memory 28 is extracted, the attribute name of the question is presented to the output unit 22, and the input is prompted. As a result, the input value from the input unit 21 is stored in the work memory 28 as a fact that the question attribute value is set.

In the above example, if "3 days" and "Kyoto" are input for "period ()" of [Question 1] and "location ()" of [Question 2], The question is translated into the following facts in the work memory 28:

[Fact 5]: Period (3) [Fact 6]: Location (Kyoto) As a result, the content in the work memory 28 has been updated, so that the process returns to the event development process S1 and continues.

[Event Expansion Step S1 (Second Time)] Since no new event has been generated, the process directly proceeds to the application rule extraction step S2.

[Applied Rule Extracting Step S2 (Second Time)] The rule 3-5 in FIG. 3 in which the condition part is satisfied by [Fact 3] in the work memory 28 and [Fact 5] newly stored is as follows:
The rule 3-1 in FIG. 3 in which the condition part is satisfied by the newly stored [Fact 6] is extracted as an applicable rule and stored in the work memory 28.

[Rule 2]: if (event name (travel) &
Period (> 1)) then Request (Accommodation); [Rule 3]: if Place (Kyoto) then Interest (Kyoto cuisine); In this way, a new applicable rule was created. I do.

[Rule execution process S3 (second time)] [Rule 2] and [Rule 3] in the work memory 28 are executed, the following facts are newly added to the work memory 28, and the process returns to the event development process S1. .

[Fact 7]: Request (accommodation) [Fact 8]: Interest (Kyoto cuisine) Here, the executed [Rule 2] and [Rule 3] do not need to be executed in the future, so the executed flag To distinguish them.

[Event Development Step S1 (Third Time)] Since no new event has been generated, the process directly proceeds to the application rule extraction step S2.

[Applied Rule Extracting Step S2 (Third)] According to [Fact 7] and [Fact 8] in the work memory 28, the rules 3-3 and 3-4 in FIG. These rules are extracted and stored in the work memory 28.

[Rule 4]: if Request (Accommodation) then
Service (Accommodation guide); [Rule 5]: if Interest (Kyoto cuisine) then Service (Kyoto cuisine guide);
Next, a rule execution step S3 is performed.

[Rule execution process S3 (third time)] [Rule 4] and [Rule 5] in the work memory 28 are executed, the following fact is newly added to the work memory 28, and the process returns to the event development process S1. .

[Fact 9]: Service (Accommodation information) [Fact 10]: Service (Kyoto cooking information) [Rule 4] and [Rule 5] executed here have been executed because they need not be executed in the future. Add flags and distinguish.

[Event Expansion Step S1 (Fourth Time)] Since no new event has been generated, the process directly proceeds to the application rule extraction step S2.

[Applied Rule Extraction Step S2 (Fourth)] Because there is no new rule that satisfies the condition part in the rule set 32 due to [Fact 1] to [Fact 10] in the work memory 28, , Applicable rules are not extracted. Further, since there is no question in the work memory 28,
Proceed to service extraction step S5.

[Service Extraction Step S5] Next, in the service extraction step S5, the fact in the work memory 28 is
Extract facts about the service. In the examples so far,
The following facts are extracted.

[Fact 4]: Service (shrine information) [Fact 9]: Service (accommodation information) [Fact 10]: Service (Kyoto cuisine information) The parameter values of the facts regarding the service obtained in this way are presented. The service is extracted as a service to be performed, and these services are presented to the output unit 22 in response to the input event. This makes it possible to present a service flexibly corresponding to the user's intention.

[0045]

As described above, according to the present invention,
If the field itself is an event that spans a plurality of different fields, the service mapping by the information retrieval unit will use the event knowledge and rule set,
Reasoning is performed using information acquisition through conversation with the user.
By linking to a service, it is possible to flexibly cope with a situation where the object of interest covers a plurality of fields, or a process of processing a user's intention in which a guidance range is different depending on viewpoints and cases.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration example of the present invention.

FIG. 2 is a diagram illustrating an example of event knowledge.

FIG. 3 is a diagram illustrating an example of a rule set.

FIG. 4 is a flowchart illustrating an information search process.

FIG. 5 is a configuration diagram of a service mapping unit in a conventional field inference method.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 11 Input part 12 Service mapping part 13 Field / service matching means 14 Field / service correspondence table 15 Output part 20 Information search part 21 Input part 22 Output part 23 Event expansion means 24 Application rule extraction means 25 Rule execution means 26 Conversation guidance means 27 Service extraction means 28 Work memory 30 Knowledge database unit 31 Event knowledge 32 Rule set

Claims (3)

[Claims] 1. An information search method for determining a service field to be searched by using a keyword representing a user intention expressed as a text as an input, the event knowledge collecting knowledge about an event corresponding to the user intention, and a condition The input keyword is collated with the event knowledge using a rule set in which rules expressed by the action part and the action part are collected.
An event expansion process for generating facts and questions and setting them in the work memory, and when a rule in the work memory that satisfies the condition part of the rule exists in the rule set,
An application rule extraction process of extracting the rules and setting them in the work memory; a rule execution process of executing the rules if the rules exist in the work memory and setting the obtained facts in the work memory; If there is a question in the memory, the user is asked a question, prompted for input, a fact is generated from the input result, and the conversation guidance process of setting it in the work memory. A service extracting step of extracting facts to be represented. 2. An information retrieval apparatus which determines a service field to be searched by using a keyword representing a user intention expressed in text as an input, and stores event knowledge obtained by collecting knowledge on an event corresponding to the user intention. Means, means for storing a rule set in which rules expressed by a condition part and an action part are stored, a work memory used for temporary storage of information, and a keyword inputted is checked against the event knowledge. Event expansion means for generating facts and questions from the knowledge of the corresponding event and setting them in the work memory, and when rules in the work memory that satisfy the condition part of the rules exist in the rule set Rule extraction means for extracting the rules and setting them in the work memory, and the extracted rules exist in the work memory A rule execution means for executing the rule and setting the obtained fact in the work memory; and, when a question exists in the work memory, asking the user,
A conversation guidance means for prompting an input, generating a fact from the input result, and setting the result in a work memory, and a service extracting means for extracting a fact representing a service field when there are no questions or rules in the work memory. Characteristic information retrieval device. 3. A storage medium storing an information search program for realizing an information search apparatus for determining a service field to be searched by using a keyword representing a user intention expressed in text as an input. Using the event knowledge that collects the knowledge about the corresponding event and the rule set that collects the rules expressed by the condition part and the action part, the input keyword is compared with the event knowledge, and the knowledge about the event corresponding to the keyword is obtained. , An event expansion process for generating facts and questions and setting the same in the work memory; and, when there is a rule in the rule set where the fact in the work memory satisfies the condition part of the rule, extracting the rule, Applying rule extraction processing to be set in the work memory and, if the extracted rule exists in the work memory, the rule Is executed, and the obtained fact is set in the work memory. If there is a question in the work memory, a question is returned to the user, prompting for input, the fact is generated from the input result, and the result is set in the work memory. Stores an information retrieval program that stores a program for causing a computer to execute a conversation guidance process and a service extraction process for extracting facts representing a service field when there are no questions or rules in the work memory. Storage media.