JP2003271615A - Answer search system, answer search method and answer search program - Google Patents

Abstract

(57) [Summary] [Problem] To promptly extract an accurate answer to a given question using a computer, and reduce the load on technical staff. SOLUTION: When a question 1 concerning various products sold to a customer is received, this is registered in a database 3 of a computer 2 and can be reused. While assigning a classification 6 or a keyword 5 to a question, a rule 10 is stored which indicates what element was used for answering the question. In the case of similar answers, their frequencies are counted up in order. The search means 15 searches the registered rules to generate an answer.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an answer search system, an answer search method, and an answer search program that can analyze and use an answer to a question from a customer regarding a product or the like.

[0002]

2. Description of the Related Art When selling a product, a sales person or a technical person receives various questions from customers. It is extremely important for business to have a system that responds appropriately to these questions. However, the number of technicians in charge of sales is insufficient in absolute numbers, and questions are concentrated on specific technicians through the sales staff. Therefore, these engineers are extremely burdened. In order to deal with this, various efforts are being made to create various manuals, educate sales personnel, and increase the number of technical personnel who respond to questions.

[0003]

By the way, the above conventional techniques have the following problems to be solved. As new products and new materials are developed, new questions are asked by customers, and it is not easy to educate salespeople and technicians to respond to these questions immediately. On the other hand, if the customer's questions are concentrated on a specific technician, the response from receiving the question to giving the answer becomes extremely poor. Also, the method of answering the question from the customer varies depending on the level of the engineer.
For example, consider the case where an Internet search engine is used to retrieve the answer corresponding to a question. Using the keywords included in the question, extract the documents that contain the keywords. However, if the number of extracted documents is large, it is not easy to select an appropriate document as an answer to the question. Therefore, a better database search engine is needed to get the best answer to the question. The present invention has been made in view of the above points, and provides an answer search system, an answer search method, and an answer search program for a question that can promptly extract an accurate answer to a given question. The purpose is to

[0004]

The present invention adopts the following constitution in order to solve the above points. <Structure 1> When an answer to a question is generated, a means for adding a classification to the question and registering it, a means for registering an element included in the answer, a classification of the question, and an element of the answer An answer search system comprising: means for registering a correlation as a rule.

The content and type of question are arbitrary. The method and contents of answering the question are also arbitrary. The format and contents of the classification given to the question are also arbitrary. The classification itself may be artificially performed by a person in charge who understands the content of the question. There are two methods for adding the question data: one is to store it together with the question data in a storage device, the other is to store the question data in an index file or the like in association with the question data. The elements included in the answer consist of materials that form the main part of the answer and information that identifies the material. The answer contains one or more elements. A rule is data in an arbitrary format that indicates the correlation between the classification of questions and the elements of answers. In addition, the rule is made to be searchable by using the question classification as a key. Searching for rules yields the element corresponding to the classification. If the rule matches the question, then that element can be used to generate the answer.

<Structure 2> In the answer search system according to Structure 1, when a classification is added to a new question, the above rule is searched, and the question and the answer element to which the corresponding classification is added are identified. An answer retrieval system comprising means for retrieving a rule indicating the relationship and a means for displaying the retrieval result.

The configuration of the search means is arbitrary. The configuration of the search means may differ depending on the rule registration method. When a new question is accepted, the category is assigned to the question by any means. Search for a rule using this classification as a key. The display method of the search result is arbitrary. You may display the rules as they are.

<Structure 3> When an answer to a question is generated, a means for registering a keyword extracted from the question, a means for registering an element included in the answer, a keyword for the question, and an element for the answer And a means for registering a correlation with as a rule, and an answer retrieval system.

The content and type of question are arbitrary. The method and contents of answering the question are also arbitrary. The format and content of the keywords extracted from the question are also arbitrary. The keyword extraction itself may be artificially performed by a person in charge who understands the content of the question. The registration method includes a method of storing the question data together with the question data in a storage device, and a method of storing the question data in association with the question data in an index file or the like. The elements included in the answer consist of materials that form the main part of the answer and information that identifies the material. The answer contains one or more elements. The rule is data in an arbitrary format that indicates the correlation between the keyword of the question and the element of the answer. Note that the rule is set so that a search can be performed using the keyword of the question as a key. When you search for a rule, you get the element that corresponds to the keyword. If the rule matches the question,
You can use that element to generate an answer.

<Structure 4> In the answer search system according to Structure 3, when a keyword is extracted from a new question, the above rule is searched, and the relationship between the question from which the corresponding keyword is extracted and the answer element. An answer retrieval system comprising means for retrieving a rule indicating and a means for displaying the retrieval result.

The structure of the search means is arbitrary. The configuration of the search means may differ depending on the rule registration method. When a new question is accepted, a keyword is extracted from the question by any means. Search for a rule using this keyword as a key. The display method of the search result is arbitrary. You may display the rules as they are.

<Structure 5> The answer search system according to any one of Structures 1 to 4, further comprising means for generating a new answer using the searched rule.

The new answer is, for example, a list of answer elements corresponding to the question according to the retrieved rule. It does not have to be something that can be submitted as an answer as it is. Anything that can be finally submitted by the person in charge is acceptable. If the elements of the answer corresponding to the question are listed according to the searched rule, it is possible to sufficiently reduce the effort of the person in charge to create the answer to the question.

<Structure 6> In the answer retrieval system according to any one of Structures 1 to 4, when a rule is used when creating an answer to a question, the frequency of occurrence of the rule is counted and recorded. An answer retrieval system characterized by having means.

There may be variations in the classification of questions and the correlation between keywords and answer elements. Therefore, the frequency of occurrence is counted and recorded. As a result, it is possible to select and use the one with a high occurrence frequency.

<Structure 7> In the reply search system according to any one of Structures 1 to 4, the reply search system is characterized in that the reply element is information indicating a storage location in which the reply is stored.

For example, an answer to a certain question can include the URL of the site displaying the reference information on the Internet. This URL can be an element of the answer. When the occurrence frequency of the URL used for a similar question is registered, the respondent can easily create a highly accurate answer.

<Structure 8> In the answer search system according to any one of Structures 1 to 4, if the classification of the question or the relationship between the keyword and the element of the answer is not registered, the work for generating the answer to the question is completed. An answer retrieval system comprising means for controlling such that it cannot be performed.

Controlling the work so that it cannot be completed means including a process of forcibly registering the classification of questions or the relationship between keywords and the elements of answers. By forcibly registering the rules, it is possible to accumulate rules and improve the accuracy of answers to questions.

<Structure 9> When an answer to a question is generated, the question is classified and registered, the elements included in the answer are registered, and the question classification and the answer element are combined. An answer search method characterized by registering correlation as a rule.

This is an answer search method corresponding to the configuration 1.

<Structure 10> In the answer search method according to Structure 9, when a new question is classified,
An answer search method characterized by searching the above rule, searching for a rule indicating a relationship between a question and an element of an answer to which a corresponding classification is given, and displaying the search result.

This is an answer search method corresponding to the configuration 2.

<Structure 11> When an answer to a question is generated, the keywords extracted from the above question are registered,
An answer search method characterized by registering an element included in an answer and registering a correlation between a keyword of the question and an element of the answer as a rule.

This is an answer search method corresponding to the configuration 3.

<Structure 12> In the answer search method according to Structure 11, when a keyword is extracted from a new question, the above rule is searched, and the relationship between the question from which the corresponding keyword is extracted and the answer element. An answer search method characterized by searching a rule that indicates and displaying the search result.

This is an answer search method corresponding to the configuration 4.

<Structure 13> In the answer search method according to any one of Structures 9 to 12, a new answer is generated using the searched rule.

This is an answer search method corresponding to the configuration 5.

<Structure 14> In the answer search method according to any one of Structures 9 to 12, when a rule is used when creating an answer to a question, the frequency of occurrence of the rule is counted and recorded. Answer search method characterized by the following.

This is an answer search method corresponding to the configuration 6.

<Structure 15> In the reply search method according to any one of Structures 9 to 12, the reply element is composed of information indicating a storage location in which the reply is stored.

This is an answer search method corresponding to the configuration 7.

<Structure 16> In the answer retrieval method according to any one of Structures 9 to 12, unless the classification of the question or the relationship between the keyword and the element of the answer is registered, the work for generating the answer to the question is completed. An answer search method that is controlled so that it cannot be done.

This is an answer search method corresponding to the structure 8.

<Structure 17> A process of adding a classification to the question and registering it when an answer to the question is generated,
An answer search program characterized by causing a computer to execute a process of registering an element included in an answer, a process of classifying the question, and a process of registering a correlation between the element of the answer as a rule.

This is an answer search program corresponding to the configuration 1.

<Structure 18> In the answer search program according to Structure 17, when a classification is added to a new question, a process for searching the above rule and the question and answer elements to which the corresponding classification is added An answer search program characterized by causing a computer to execute a process of searching for a rule indicating a relationship with and a process of displaying the search result.

This is an answer search program corresponding to the configuration 2.

<Structure 19> When an answer to a question is generated, a process of registering a keyword extracted from the question, a process of registering an element included in the answer, a keyword of the question, and an element of the answer An answer search program characterized by causing a computer to execute a process of registering a correlation with a rule as a rule.

An answer search program corresponding to the configuration 3.

<Structure 20> In the answer search program according to Structure 19, when a keyword is extracted from a new question, a process for searching the rule and a question and an answer element from which the corresponding keyword is extracted are provided. An answer search program characterized by causing a computer to execute a process of searching a rule indicating a relationship and a process of displaying the search result.

This is an answer search program corresponding to the configuration 4.

<Structure 21> In the answer search program according to any one of Structures 17 to 20, an answer characterized by causing a computer to execute a process of generating a new answer using the searched rule Search program.

An answer search program corresponding to the configuration 5.

<Structure 22> In the answer search program according to any one of Structures 17 to 20, when a rule is used when an answer is made to a question, the frequency of occurrence of the rule is counted and recorded. An answer search program characterized by causing a computer to execute the processing.

This is an answer search program corresponding to the structure 6.

<Structure 23> In the reply search program according to any one of Structures 17 to 20, the reply element is
An answer search program characterized by comprising information indicating a storage location for storing answers.

This is an answer search program corresponding to the configuration 7.

<Structure 24> In the answer retrieval program according to any one of Structures 17 to 20, unless the classification of the question or the relationship between the keyword and the element of the answer is registered, the work for generating the answer to the question is completed. An answer search program characterized by causing a computer to execute a process of controlling so that it cannot be performed.

This is an answer search program corresponding to the structure 8.

[0052]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to specific examples. FIG. 1 is a block diagram showing a specific example of the answer search system of the present invention. In this system, when a question 1 about various products sold to a customer is received, it is registered in the database 3 of the computer 2 and can be reused. A network 4 is connected to the computer 2, and the question 1 is taken in through the network 4. Of course, it may be imported from another route. The database 3 includes information such as the question 1 to which the classification 6 is added as shown in the figure, the rule 10 described later, the element 8, the set of the question 1 and the answer 7, and the like.

Question 1 received from a customer or the like often has similar contents. The question 1 is assigned the classification 6 and stored in the database 3. Also, if there is a keyword 5 that can identify the gist of question 1, along with classification 6,
Alternatively, instead of the category 6, the question 1 is assigned and stored in the database 3. For questions, process in this way,
Answer 7 is processed by another method.

The answer 7 to the question 1 varies depending on the content of the question 1. However, all answers 7 must include elements 8-1 to 8-3 such as data analysis results, specifications, operation manuals, links with reference documents, characteristic lists, and price lists. Contains. Although only three types of elements are shown in this figure, the number of elements is arbitrary. An element refers to the information presented to specifically answer a question or the document itself. Also, or alternatively, the element may be information indicating the location of the reference material. That is, in general, the answer is made up of one or more of these elements.
Therefore, while assigning the category 6 or the keyword 5 to the question, the results of what elements are used in the answer to the question are accumulated.

The information to be accumulated is the information indicating the correlation between the classification given to the question and the answer element, or the correlation between the keyword included in the question and the answer element. This correlation is called a rule. Different respondents to the same question may have slightly different elements of the answer, but similar questions are often answered using similar elements. Therefore, the frequently-used correlation among the question and answer elements is stored as the rule 10. By accumulating the correlation between a large number of questions and the answers to the questions in the manner described above and creating a rule, it is possible to find out the regularity that it is sufficient to take out the element and answer with this rule for this question. The search means 15 has a function of searching a registered rule and generating an answer.

FIG. 2 is an explanatory diagram showing an example of specific questions and answers. In this figure, question 1 is displayed in the upper half and answer 7 is displayed in the lower half. This data is directly stored in the database 3 shown in FIG. 1 as history data.
Using such data, data such as rules to be described later is created. For example, "Semiconductor AAXXX
In case of the question "Please tell me the temperature characteristics of", the classification is "semiconductor" and the keyword is the semiconductor name "AAXX
"X" and "temperature characteristic". Answer 7 to this is
"Semiconductor AXXX has the types shown in Specification No. XXX. The temperature characteristics of each are listed in Table N.
o. As shown in XXX. It becomes the contents such as ". At this time, the element of answer 7 is the specification number. XXX and list No. It is XXX.

When the question is accepted, the classification 6-1 and the keywords 5-1 and 5-2 are input as shown in the figure. The input processing such as the classification may be performed by the person in charge of accepting the question or by the respondent who created the answer. When the answer 7 is created, the respondent inputs the answer elements 10-1 and 10-2. At this time, the result that the element 10-1 is used is input as an answer to the combination of the classification 6-1 and the keyword 5-1. This is registered as a rule. Similarly for the element 10-2, the result that this element is used as an answer to the classification 6-1 and the keyword 5-2 is input.

FIG. 3 is an explanatory diagram showing a specific example of the rule. Based on the relationship between question 1 and answer 7 shown in FIG.
The rule shown in (a) is obtained. Here, the most specific of the obtained rules is (semiconductor & name AAXXX)
-(Specification No. XXX), (Semiconductor & temperature characteristics)-
(List No. XXX). In addition, (Semiconductor)-(Specification No. XXX), which is a little abstract,
(Semiconductor)-(List No. XXX), (Temperature Characteristics)-
Rules such as (List No. XXX) are also possible. While the relationship between these questions and answers is accumulated as a rule, the frequency of occurrence of that rule is counted and recorded.

If the same rule is applied to similar questions, it is a highly accurate rule and an answer to the question can be generated almost automatically. The generated answer may be evaluated and returned by an engineer. The rules may change over time. For example, when a certain rule is established once, a point of 1 is counted. The point is the frequency of occurrence. The rule with high occurrence frequency has high accuracy. Rules are automatically generated, but can be re-evaluated and modified by experts to improve quality.

The answer element is the URL of the Internet
(Uniform Resource Locator) may be included. For example, in response to a question, the URL adopted most frequently can be the element of the correct answer. For any kind of question, you can answer which site is best to refer to, which folder in the database is best to refer to.

The rule shown in FIG. 3 (b) is a rule in which the URL used for the answer is associated with the keyword included in the question. When many sites are used for the same question, count the frequency of occurrence and keep a record of which site was used most in response to this question. This makes it possible to apply rules and automatically include the most frequently occurring sites in the answers when similar questions are asked.

FIG. 4 is a flowchart showing the specific processing contents of the workflow system. In order to operate the above system, it is necessary to classify questions in the database and register keywords, answer elements, and the like. By performing this work for every question as much as possible, a highly reliable database can be constructed. For this purpose, a workflow system may be used for the registration process. It is better to take in. The workflow system has a function of automatically and forcibly registering necessary items in the database according to the progress of work.

In the flow shown in FIG. 4, first, step S1
At, the question acceptance process from the customer is executed. This is a process of registering the question content as shown in FIG. 2 as data. After that, in step S2, a salesperson who processes this question is designated, and the salesperson inputs the examination result for the question. This is what you do when you have a question from an important customer. Step S2 is not necessary when simply replying by e-mail using the network. In this case, in step S2, the category of the question and the keyword are input as shown in FIG. In addition, decide which technical contact to ask the question. If the sales representative is not decided, these steps are performed in step S.
Done in 3.

At the next step S3, the technical person handed over the question creates and inputs an answer. The contents of the answer are as shown in FIG. Furthermore, the elements of the answer are also input in the manner shown in FIG. In the next step S4, a circulation process for approval of the answer by the boss, manager, and other related persons is executed. To the people concerned,
This is a process in which the contents of the question and the answer are circulated as they are, and the approval is checked. When these processes are completed, a database registration process of question and answer contents is performed in step S5.

Here, the contents of the question and answer and the rules as shown in FIG. 1 are registered. First, in step S6, the set of questions and answers is directly registered as a data file. Next, in step S7, the answer is decomposed into elements,
If you specify an element, it will be stored in the database for retrieval. That is, when a certain element is specified as an answer to a certain question, it can be immediately retrieved from the database. Then, step S8
In step 3, the question and answer data are analyzed to create a rule as described with reference to FIG.

Next, in step S9, the frequency of the rule is registered. With a new rule, the frequency is 1. If you are using the rules you have already used in the past,
The frequency of the corresponding rule is incremented by 1. When such a database registration process is completed, the process proceeds to step S10 and the reply is returned to the sales person. Finally step S11
At, the salesperson executes the process of transmitting the answer to the customer. In the case of processing in which the sales representative is not determined, the step S11 is skipped and the technical representative directly sends the reply to the customer. Using the workflow shown in this figure, the rules corresponding to the above questions and answers are sequentially accumulated, the frequency is calculated for each rule,
A highly practical database is created.

FIG. 5 is a flowchart of a more specific processing example when utilizing the system of the present invention. In the system as shown in FIG. 1, first, when a question comes, the question is classified in step S21. Next, in step S22, keywords are extracted from this question. For example, when the question 1 as shown in FIG. 2 is received, the question is classified into semiconductors, and the name AAXXX of the semiconductor and the temperature characteristic which is the keyword are extracted from the question. Next, in step S23,
Search for rules stored in the database.

In this case, "semiconductor" or "semiconductor AA"
A rule is searched using a keyword such as "XXX" or a temperature characteristic as a key. As a result, as shown in the figure, in the past example, the rule that the specifications AAXXX and the list AAXXX are used as the answers is extracted. When some rules are extracted and some answer elements are collected, an appropriate one is selected from these answer elements (step S24). Then, in step S25, the answer is edited. Up to this part, almost all can be automatically processed by a computer.

The answer element may be selected by the person in charge. Then, in step S26, the result is sent to an appropriate technical person. The technician will now read both the answer and the question to determine if it is appropriate. This allows
The technical person can process the reply to the approval route in an extremely short time. That is, since the answer is not created from the beginning as in the past, the engineer is not burdened with much. When the answer is determined in this way, the process proceeds to step S4 and subsequent steps in FIG. Then, in the case of similar answers, the frequency is sequentially counted up. Therefore, when a large number of similar questions are received, a mechanism for automatically obtaining an optimum answer is created.

As described above, it is possible to provide an environment in which an appropriate answer can be selected by using the rules in the database and the answer can be returned as an answer to many customers' questions without trouble. These rules can also be used to educate salespeople. The present invention is not limited to the above embodiments. In the above example, the method of answering the customer's question regarding the product has been described, but the type of question is arbitrary. In addition, you can freely divide the answer into different elements and how to classify the question. That is, it can be widely used as a search system for extracting an answer corresponding to a question in various fields.

Each functional block shown in FIG. 1 may be composed of a separate program module or an integrated program module. Further, all or part of these functional blocks may be configured by hardware by a logic circuit.
Further, each program module may be incorporated into an existing application program to be operated, or may be operated as an independent program. The computer program for implementing the present invention as described above can be recorded in a computer-readable recording medium such as a CD-ROM, installed, and used. It can also be downloaded and used in the memory of a computer via a network.

[Brief description of drawings]

FIG. 1 is a block diagram showing a specific example of an answer search system of the present invention.

FIG. 2 is an explanatory diagram showing an example of specific questions and answers.

FIG. 3 is an explanatory diagram showing a specific example of a rule.

FIG. 4 is a flowchart showing specific processing contents of the workflow system.

FIG. 5 is a flowchart of a more specific processing example when utilizing the system of the present invention.

[Explanation of symbols]

1 question 2 computers 3 database 4 network 5 keywords 6 classification 7 answers 8-1, 8-2, 8-3 elements 10 rules

Claims (24)

[Claims] 1. When an answer to a question is generated,
A unit for adding a classification to the question and registering; a unit for registering an element included in an answer; a unit for registering the correlation between the question classification and the element of the answer as a rule. Answer search system characterized by. 2. The answer retrieval system according to claim 1, wherein when a classification is given to a new question, the rule is searched to find the question and the element of the answer to which the corresponding classification is given. An answer search system comprising means for searching a rule indicating a relationship and means for displaying the search result. 3. When an answer to a question is generated,
Means for registering a keyword extracted from the question; means for registering an element included in the answer; means for registering a correlation between the keyword of the question and the element of the answer as a rule, Characteristic answer search system. 4. The answer search system according to claim 3, wherein when a keyword is extracted from a new question, the rule is searched to find the relationship between the question from which the corresponding keyword is extracted and the answer element. An answer retrieval system comprising means for retrieving the indicated rule and means for displaying the retrieval result. 5. The answer search system according to claim 1, further comprising: a means for generating a new answer using the searched rule. 6. The answer search system according to claim 1, wherein when a rule is used when creating an answer to a question, a frequency of occurrence of the rule is counted and recorded. An answer search system characterized by having. 7. The answer search system according to claim 1, wherein an element of the answer is information indicating a storage location in which the answer is stored. 8. The answer search system according to any one of claims 1 to 4, unless the classification of the question or the relationship between the keyword and the element of the answer is registered, the work of generating the answer to the question cannot be completed. An answer retrieval system comprising means for controlling as described above. 9. When an answer to a question is generated,
An answer search method, characterized in that the question is classified and registered, the elements included in the answer are registered, and the correlation between the question classification and the answer element is registered as a rule. 10. The answer search method according to claim 9, wherein when a classification is given to a new question, the rule is searched, and the question and the element of the answer to which the corresponding classification is given are classified. An answer search method characterized by searching a rule indicating a relationship and displaying the search result. 11. When an answer to a question is generated, a keyword extracted from the question is registered, elements included in the answer are registered, and the correlation between the keyword of the question and the element of the answer is registered. An answer search method characterized by being registered as a rule. 12. The answer search method according to claim 11, wherein when a keyword is extracted from a new question, the rule is searched to find the relationship between the question from which the corresponding keyword is extracted and the answer element. An answer search method characterized by searching the rule shown and displaying the search result. 13. The answer search method according to claim 9, wherein a new answer is generated using the searched rule. 14. The answer search method according to claim 9, wherein when a rule is used when creating an answer to a question, the frequency of occurrence of the rule is counted and recorded. Answer search method characterized by. 15. The reply search method according to claim 9, wherein the reply element comprises information indicating a storage location in which the reply is stored. 16. The answer search method according to claim 9, wherein the task of generating an answer to the question cannot be completed unless the classification of the question or the relationship between the keyword and the element of the answer is registered. An answer search method characterized by controlling as described above. 17. When an answer to a question is generated, a process of adding a classification to the question and registering it, a process of registering an element included in the answer, a classification of the question, and an element of the answer An answer search program characterized by causing a computer to execute the process of registering the correlation of as a rule. 18. The answer search program according to claim 17, wherein when a classification is given to a new question, a process of searching the rule, and a question and an answer element to which the corresponding classification is given An answer search program characterized by causing a computer to execute a process of searching for a rule indicating the relationship between and a process of displaying the search result. 19. When an answer to a question is generated, a process of registering a keyword extracted from the question, a process of registering an element included in the answer, a keyword of the question, and an element of the answer An answer search program characterized by causing a computer to execute a process of registering a correlation as a rule. 20. The answer search program according to claim 19, wherein when a keyword is extracted from a new question, a process of searching the rule and a relationship between the question from which the corresponding keyword is extracted and the answer element An answer search program characterized by causing a computer to execute a process of searching for a rule that indicates and a process of displaying the search result. 21. The answer search program according to claim 17, further comprising causing a computer to execute a process of generating a new answer using the searched rule. program. 22. In the answer search program according to claim 17, when a rule is used when creating an answer to a question, the frequency of occurrence of the rule is counted and recorded. An answer search program characterized by causing a computer to execute processing. 23. The answer search program according to claim 17, wherein an element of the answer is information indicating a storage location in which the answer is stored. 24. In the answer search program according to claim 17, the work of generating an answer to a question cannot be completed unless the classification of the question or the relationship between the keyword and the element of the answer is registered. An answer search program characterized by causing a computer to execute the process of controlling as described above.