JP2001134590A - Knowledge management and retrieval system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a knowledge management and retrieval system which is easy for a user to use by actualizing both the improvement of retrieval precision and the effective use of information. SOLUTION: This knowledge management and retrieval system decides whether Japanese document data of user's electronic mail are deficient or have an error by comparing the Japanese document data with a standard document structure model, automatically generates and sends back to the user electronic mail making a request to compensate the deficiency of the Japanese document data or correct the error when deciding that the Japanese document data are deficient or have the error, and generates accurate Japanese document data by compensation and correction using Japanese document data sent by from the user by electronic mail.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention analyzes the contents of an e-mail about a question or question transmitted from a user via the Internet, and returns an answer searched and extracted from a database to the user based on the analysis result. And a knowledge management and retrieval system that responds to user questions and questions.

[0002]

2. Description of the Related Art With the rapid spread of the Internet, an environment in which business people and ordinary people can use enormous amounts of information is being prepared. In order to cope with such an environment, a data mining tool (Data
Mining Tool) and a Knowledge Management System, which allows many people to share information collectively and utilize information as a whole, are being developed.

Further, development of a system (hereinafter, referred to as a knowledge management / retrieval system) combining such a data mining tool and a knowledge management system is being studied. This knowledge management and search system analyzes the contents of e-mails in Japanese sentences about questions and questions sent from users via the Internet, searches the database for desired information from the user, and returns search results to the user. User questions
It is a system that answers questions.

[0004]

Such a data mining tool, knowledge management system and knowledge management / search system perform management / search using Japanese document data described in Japanese. However, among the data mining tools and knowledge management systems described above, with regard to data mining tools and knowledge management systems for Japanese document data, there are few products suitable for widespread use at present.

[0005] The problem with current data mining tools and knowledge management systems for Japanese document data is that they are intended to be versatile and to deal with vast amounts of information in a wide category without limiting the information. The data mining tool is time-consuming to search, and cannot utilize the knowledge management system efficiently.

It is expected that the performance of a knowledge management / search system combining such a data mining tool and a knowledge management system cannot be expected as it is. There was a request to break through these situations and create a knowledge management and search system that would enable both quick search and efficient use of information.

Further, as a problem when operating the knowledge management / retrieval system, there is a countermeasure against a frequently asked question (FAQ) mail concerning a question / question transmitted from a user. FAQ mail is a question or question of the same sentence sent from many users, and it is laborious to repeat the same answer to FAQ mail.
A response mail for AQ mail is prepared in advance and stored in a database, the contents of the transmitted FAQ mail are analyzed, a response is searched and extracted from the database, and the response mail is automatically returned to the user. A knowledge management and search system is being considered.

[0008] However, it is expected that the answers to these FAQ mails will not be answers that will allow the user to answer the questions / questions. This is because the information described in the FAQ mail transmitted from the user may be incorrect or missing, and the accuracy of the search is low even if analysis is performed using such FAQ mail. The user may not be able to obtain a desired answer and may repeat the search operation. For this reason, there is a possibility that a knowledge management and search system that cannot quickly search and efficiently utilize information may be used.

[0009] Even if there is no error or lack of information described in the FAQ mail transmitted from the user, if there is a problem in the analysis itself, the user will not be able to answer the question or answer the question. . It is also necessary to employ an analysis algorithm that is devised to improve the accuracy of the search. Furthermore, even if the analysis algorithm is devised, it may still be possible to search and extract answers that have not achieved a target.
It is also necessary to eliminate unanswered answers so as to increase the reliability of general users in the system.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-described problems, and provides a knowledge management and search system which achieves both improvement in search accuracy and efficient use of information, and which is easy to use for users. The purpose is to do.

[0011]

According to a first aspect of the present invention, there is provided a knowledge management and retrieval system.
Respond to users' questions and questions by analyzing the contents of e-mails about questions and questions sent from users, searching the knowledge database based on the analysis results, extracting answers, and returning the answers to the users. In the knowledge management and search system, the Japanese document data of the user's e-mail is compared with the standard document structure model to determine whether or not the Japanese document data is missing or incorrect. If it is determined that the data is missing or incorrect, an e-mail is automatically sent to the user requesting that the Japanese sentence data be supplemented or corrected, and the e-mail sent from the user is returned. It is necessary to provide control means for performing an information correction procedure for creating correct Japanese document data by supplementing and correcting using Japanese sentence data of mail. And butterflies.

According to a second aspect of the present invention, in the knowledge management / search system according to the first aspect, the information correction step counts the number of times of e-mail replies to the user, If the number of replies reaches a predetermined number, instead of automatically creating and replying to the e-mail, create an e-mail to request that the operator supplement the missing Japanese text data or correct the error by manual input by the operator. And sending a reply.

According to the third aspect of the present invention, the content of the e-mail about the question / question transmitted from the user is analyzed, and the knowledge database is searched based on the analysis result and the answer is obtained. In a knowledge management and search system that responds to user questions and questions by extracting and returning answers to the user, search and extract keywords related to specific information from Japanese document data of e-mail from the user, Each knowledge database that is classified for each accesses a knowledge database group in which a plurality of knowledge databases are recorded, compares the category classification keyword assigned to each of the plurality of knowledge databases with the keyword retrieved and extracted earlier, and receives an e-mail from the user. A knowledge database that matches the information category of Japanese document data A control means for performing categorization procedure to choose from, characterized in that it comprises.

According to a fourth aspect of the present invention, in the knowledge management / search system according to the third aspect, the categorizing step is performed based on knowledge data extracted from a knowledge database as a result of answer search. Searching and extracting keywords related to specific information, accessing a knowledge database group in which multiple knowledge databases classified for each information category are recorded, and category classification keywords assigned to each of the multiple knowledge databases and previously searched and extracted keywords And select a knowledge database that matches the information category of the Japanese document data of the e-mail from the user from the group of knowledge databases, and add the previously searched and extracted knowledge data to the selected knowledge database to create a database. It is characterized by including the procedure to update

According to the knowledge management / search system of the present invention, the contents of the e-mail about the question / question transmitted from the user are analyzed, and the knowledge database is searched based on the analysis result to answer. In a knowledge management and search system that responds to a user's question or question by extracting and replying an answer to the user, the syntax structure of the knowledge data extracted from the database as a result of the search is compared with various sample data. And a control unit for performing a procedure for improving the accuracy of answer contents by selecting and extracting the matching knowledge data.

According to the knowledge management / retrieval system according to the sixth aspect, in the knowledge management / retrieval system according to the fifth aspect, the procedure for improving the accuracy of the answer content is performed by using the knowledge data extracted from the database as a result of the search. To check whether there is a problem with the syntactic structure, check whether the subject matter is reverse, check the pattern on the syntactic structure, and perform multiple or all of these checks. The method includes a step of analyzing and selecting and extracting knowledge data.

[0017]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a knowledge management / search system according to the present invention will be described. In the description of the present embodiment,
A description will be given assuming a medical field as a specific example of knowledge to be managed. FIG. 1 is an explanatory diagram for explaining an overall system for operating this embodiment, and FIG. 2 is an explanatory diagram for explaining the concept of a knowledge management / search system.

As shown in FIG. 1, the user and the operator of the knowledge management / retrieval system are connected via a telecommunication facility 1 such as the Internet. A user terminal 2 is provided on the user side to connect with the telecommunications equipment 1, and a communication management server 3, an e-mail server 4, a knowledge management / search system operation server 5, and a database are provided on the knowledge management / search system operator side. A server 6 and an operator terminal 7 are provided.

In this specification, the telecommunications equipment 1 is an Internet network, which collectively refers to equipment for connecting data terminals. For example, if the telecommunications equipment 1 to which the user terminal 2 is connected is a packet switched network among public lines, the telecommunications equipment 1 is a DSU (Digital Service Unit).
Will be included. Further, the telecommunications equipment 1 to which the user terminal 2 is connected is an ISDN (Integrated
Services Digital Network)
Includes DSU and TA (Terminal Adapter).

If the telecommunications equipment 1 to which the user terminal 2 is connected is an analog network of a public line, the telecommunications equipment 1 includes a modem. Note that the user terminal 2 may have a built-in modem. When the user terminal 2 is connected to a local facility such as a LAN (Local Area Network), the user terminal 2 is connected to the LAN.
Also includes equipment such as gateways and routers that make up the system.

As described above, the telecommunications facility 1 can use various telecommunications facilities capable of transmitting electronic mail from the user terminal 2 to the communication server 3. In the description of the present embodiment, the description will be made on the assumption that the Internet network uses a public line.

The user terminal 2 operated by the user is an Internet-only terminal device / personal computer (P
C) Portable information terminal device. Internet-only terminals are available for those who want to use the Internet
Developed with consideration for those who find it difficult to operate C,
A terminal device that can easily access the Internet and send and receive e-mails using a television as a display. The PC is a normal desktop laptop personal computer with internet access.

A portable information terminal device is a PDA (Personal
Digital Assistant) An information terminal device such as a palmtop personal computer. The user terminal 2 is a concept that collectively refers to these Internet-dedicated terminal devices, PCs, portable information terminal devices, and the like, and is connected to the telecommunications facility 1 via a built-in / external modem as described above. You.

The operator of the knowledge management / retrieval system is LA
In FIG. 1, the communication server 3, the e-mail server 4, the knowledge management / search system operation server 5, the database server 6, and the operator terminal 7 are connected to one cable in FIG. A bus-type network is shown. Incidentally, although not shown, it goes without saying that a ring-type / hub-type network type may be used.

The communication server 3 communicates with the telecommunications equipment 1 and the LA
Since the communication protocol differs between N and N, it has a function of converting the data into communication data corresponding to each communication protocol and outputting the data. The e-mail server 4 uses the SMTP
(Simple Male Transfer Protocol) or other mail data is transmitted and received between the communication server 3 and the knowledge management / search system operation server 5.

The knowledge management / search system operation server 5
The contents of the mail are analyzed, knowledge data corresponding to the answer is extracted from the database server 6 based on the analysis result, the knowledge data is processed, and the e-mail is sent to the e-mail server 4, the communication server 3, and the electric communication. The data is transmitted to the user terminal 2 via the facility 1. The details of the content analysis will be described later.

The database server 6 has a knowledge database group including a plurality of knowledge databases classified according to predetermined information categories. A category classification keyword is assigned to each knowledge database in order to facilitate the classification of the knowledge database, and can be used when the knowledge management / search system operation server 5 searches and extracts desired knowledge data.

The operator terminal 7 is provided for the operator to monitor the operation state of the knowledge management / search system.
The extracted knowledge data is monitored by a human, and the knowledge data is deleted if the answer is considered to be an incorrect search. Also, depending on the operation proficiency of the user who operates the user terminal 2, the user is forced to perform the operation repeatedly, and the operator assists or assists such a user (specifically, a desired answer is given). Sending e-mail to derive). The overall system for using and operating the knowledge management / search system is configured as described above.

Next, the specific operation of each unit in such a knowledge management / search system will be described. As shown in FIG. 2, the user transmits, for example, an e-mail regarding the question / question specified in the medical field from the user terminal 2.
In this case, it is assumed that the user knows that this is a knowledge management / search system that is operated only in the medical field, and does not send e-mails regarding questions / questions in fields other than the medical field.

This e-mail is automatically transmitted to the knowledge management / search system operation server 5 via the communication server 3. In the knowledge management / search system operation server 5, H
Japanese document data, which is a natural language, is extracted from e-mail data described in TML (Hyper Text Makeup Language) or text format, excluding the control language. The extraction is performed by referring to the JIS code or the like.

Subsequently, the knowledge management / search system operation server 5 performs a search using the Japanese document data. Note that the knowledge management / search system operation server 5
It corresponds to the control means of the present invention, and includes various engines and filters as shown in FIG. Further, various engines and filters indicate algorithms for realizing specific functions, and will be described as corresponding to the procedure of the present invention.

For this Japanese document data, first, an information correction engine 10 that supplements and corrects the missing information (hereinafter, the complement and correction are collectively corrected) performs an information correction operation on the Japanese text data. . The Japanese sentence data to be corrected is the title and the text of the e-mail. First, the Japanese sentence data and the syntax structure are checked for deficiencies and errors.

In the case of Japanese sentence data such as "What kind of attention is needed in daily life because of high blood pressure?", It is determined that there is no problem in the syntax structure, and the process exits the information correction engine 10. Become. However, suppose that "It is expensive, so what kind of attention is needed in life?"
If there is an obvious lack of input, it will analyze the syntactic structure, determine that the subject is missing, and point out any questions (for example,
"Please re-enter what is expensive."
Is created, and the electronic mail is returned to the user terminal 2 to urge the user to input again.

Since the user replies an answer such as "blood pressure." By e-mail, the user's question / question is based on a combination of multiple times Japanese document data from the user. Is the information correction engine 10
Recognize by By using such an information correction engine 10, the accuracy of information to be searched can be improved, and as a result, information desired by the user can be quickly obtained.

It should be noted that a user who is not accustomed to operating the user terminal 2 such as a PC or a middle-aged or elderly user may not be able to input an answer accurately. If the information correction engine 10 repeatedly inquires such a user for an infinite number of inputs, the user may decide that the Internet is still difficult, and may avoid using the system. Knowledge management available
Contrary to the intent of providing a search system.

For this reason, if the inquiry continues more than a predetermined number of times, the information correction engine 10 of the knowledge management / search system operation server 5 switches from automatic mail reply to manual mail reply by the operator terminal 7. Control. The manual operation of the operator who operates the operator terminal 7 guides the user to provide the missing information accurately. Since the operator requests that only the missing information be returned, the user need only input a minimum amount of information, and the burden on the user is reduced. Such information correction engine 10 avoids a situation in which the information necessary for the search is insufficient, or a search using erroneous Japanese document data.

Subsequently, the knowledge database is selected using the categorizing filter 20 for the original Japanese document data or the Japanese sentence data corrected by the information correction engine 10. Categorize filter 20
Then, a keyword related to the medical field is searched and extracted from the Japanese document data of the e-mail from the user, and a knowledge database that matches the information category of the medical field related to the user's question / question is selected from the knowledge database group. Controls the database server 6.

Here, if the knowledge database relates to the medical field, it is classified into information categories such as "blood pressure", "cancer" and "stress", and a knowledge database group which is a set of these knowledge databases is a database server. 6 is stored in a large-capacity hard disk or the like.

Further, a category classification keyword used for search is assigned to each knowledge database group of the database server 6. As an example of the category classification keyword, the knowledge database of “blood pressure” includes various data related to “blood pressure” such as “blood pressure”, “high blood pressure”, “low blood pressure”, “salt”, “anemia”,. Keywords are given. The categorizing filter 20 selects a knowledge database using these category classification keywords.

For example, if the e-mail from the user is Japanese sentence data such as "What kind of attention is needed in life because blood pressure is high?", "Blood pressure", "high", "life", Keywords such as "attention" and "necessary" are selected and extracted, and these keywords are compared with the category classification keywords, and a knowledge database with matching keywords is selected. The categorizing filter 20 selects and restricts the knowledge database, so that the subsequent search becomes quick and easy.

Subsequently, the search engine 30 searches the plurality of knowledge databases selected by the categorizing filter 20 using the keyword. Here, the term "search" refers to extracting relevant knowledge data from a group of knowledge data (hereinafter simply referred to as "knowledge data") of Japanese sentences registered in the knowledge database.

The search engine 30 is an engine including a search engine 30a and a concept dictionary database 30b. The search engine 30a first accesses the concept dictionary database 30b and extracts keywords, synonyms and synonyms. For example, “blood pressure” / “high” and “hypertension” are synonyms. Keywords including such synonyms and synonyms are extracted from the concept dictionary database 30b to prepare a keyword group.

Also, "blood pressure", "high", "life",
Weighting is performed with emphasis on the meaning connection between the keywords "caution" and "necessary", and the search accuracy is further improved. For example, even if there is a word such as “high living level” in the knowledge data, it is contrary to the purpose requested by the user, and “high” and “life” have low meaning connection between keywords. A sentence in which "high" and "blood pressure" appear in the same sentence is retrieved and extracted, and a sentence in which "high" and "life" appear in the same sentence are not extracted. In this way, the significance of the meaning connection between the keywords is registered as weight data in the concept dictionary database 30b.

As described above, “blood pressure” / “high” and “hypertension” are synonyms, and “blood pressure” is “high” / “high”.
The fact that the weight of the connection with "low" is high is stored in the weight data concept dictionary database 30b, so that it is possible to extract even if the word is different, and it is possible to avoid extracting sentence data having a different meaning, thereby improving search accuracy. Will be higher. Using such a keyword group and the weight data between keywords, a sentence of all the knowledge data registered in the knowledge database is searched in a brute force manner, a knowledge data including the keyword group is selected, and the weight data is further searched. In consideration of the above, the relevant knowledge data is selected and extracted. The search engine 30 performs such processing.

Although the amount of data stored in the concept dictionary database 30b is enormous, it can be stored in a large-capacity server in recent years because the data is stored in a limited field such as medical care. Furthermore, since the concept dictionary database 30b is updated as needed as the present system is operated, the accuracy can be improved.

After the search by the search engine 30 is completed, the categorization filter 20 performs another categorization. The categorization after the search is to update the knowledge database. Specifically, keywords related to the user's question / question are searched and extracted from the knowledge data obtained after the search, and the category classification keywords assigned to each knowledge database are compared with the previously searched and extracted keywords. Is to select a knowledge database that matches the information category of the e-mail from the group of knowledge databases, add and register knowledge data in the selected knowledge database, and update the database.

If a knowledge database different from the knowledge database to which the user belongs to is detected by the categorizing filter 20, the knowledge data is written in a different knowledge database to update the knowledge database. Knowledge database is categorized filter 20
As the system operation progresses, the amount of knowledge data registered in the knowledge database increases, and the knowledge database shifts so as to further increase the search accuracy.

Now, the answer content accuracy improving engine 40 for the plurality of knowledge data searched by the search engine 30 is described.
The processing is performed by The answer content accuracy improvement engine 40 will be described. Simple theory check, combination theory check, negative form check,
A pattern check is performed.

The simple theory check will be described. For example, if the previous question "What kind of life do you need to pay attention to because your blood pressure is high?"
The words "exercise" and "stress" are expected to be included in the answer, and words such as "strive nourishment and strength" and "fracture" cannot be expected. This is a theoretical check in which logical forms normally exchanged in the medical field are collected in advance and an answer suitable for a question to the user is extracted.

Next, the combination theory check will be described. The combination theory check is a theory check in which a simple theory check for checking one keyword as described above is performed by using a plurality of keywords together to further improve accuracy.

Next, the negative form theory check will be described. The negation logic check extracts a paradoxical sentence / phrase. For example, the paradox of "need to exercise" such as "needs recuperation" and "do not exercise" cannot be expected as an answer to prevent hypertension. In order to detect such a paradox, it is a logic check to detect a sentence that may have a different meaning by detecting an antonym or a negative form in a sentence or phrase.

Next, the pattern check will be described. In the pattern check, it is checked whether or not it matches the pre-patterned answer contents such as "need attention" so as to respond to the question "needs attention", and it differs from the pattern In such a case, it is a logical check to convert the answer so as to match the pattern.

As described above, the answer content accuracy improvement engine 4
In the case of 0, these simple logic check, combination logic check, negative logic check and pattern check are applied to the knowledge data to further improve the accuracy of the answer. The operator also checks the simple logic check / combination logic check / negative logic check and the pattern check through the operator terminal 7. If a correct answer is deleted by mistake, the answer is restored. When the answer is output, the answer is deleted. By performing such operations, search accuracy is further improved.

The expression conversion engine 50 performs processing on the knowledge data of the search result whose accuracy has been improved by the answer content accuracy improvement engine 40. For example, since synonyms and synonyms are also used as keywords, search and extraction may be performed using, for example, the keyword “hypertension”, which is a medical term of “hypertension”. The use of medical terms makes it difficult for users other than medical personnel to understand. In such a case, "hypertension" of the retrieved and extracted knowledge data is converted to "hypertension". In this way, the expression conversion engine 50 selects a term that is difficult to understand, such as a technical term, converts it into a general term, and outputs it.

In some cases, certain words such as "high blood pressure" are used in the knowledge data, but they have a somewhat high-pressure feeling, which is not preferable. Therefore, in order to make these assertive words such as "..da" and "is." Convert to respected words. In this way, the expression conversion engine 50 determines an optimal expression answer.

The knowledge management / search system operation server 5
At least these information correction engine 10, categorization filter 20, search engine 30, answer content accuracy improvement engine 40, and expression conversion engine 50 search and extract the most suitable knowledge data to solve the user's question / question, and obtain this knowledge. An e-mail will be output with the data as the answer.

In addition to these filters and engines,
The processing may be performed using the interest recognition engine 60. The interest recognition engine 60 processes the Japanese sentence data obtained by the information correction engine 10 in the question-and-answer session with the user, constructs user model data indicating the user's interest, and records it as a history. Then, at another opportunity, the information correction engine 1
When "0" is used, an inquiry is made also about items that the user has shown interest in the past using the user model data, and information that the user has forgotten to write in the e-mail is heard. For example, information on past histories is listed so that the user can select it.

For example, when a question / question about “blood pressure” is sent, if it is found from the user model data that there is a question and answer on “headache” in the past, the user is asked “question about headache in the past. Yes, but what about this time? "
Is obtained, the keywords "blood pressure" and "headache" can be obtained. If the search engine 30 performs processing using these keywords, the search accuracy will increase.

Further, when using the categorizing filter 20 and the search engine 30, it is possible to use the user model data to add, as keywords, items that the user has shown interest in the past. As described above, the interest recognition engine 60 allows the user to use items that the user has expressed interest in the past, and obtains a result that matches the interests of the user.

As described above, in the knowledge management / search system of the present embodiment, the information correction engine 10, the categorization filter 20, the search engine 30, the answer content accuracy improvement engine 40, the expression conversion engine 50, and the interest recognition engine 60 are used. By analyzing the contents of the e-mail about questions and questions sent from the user, searching the knowledge database based on the analysis results, extracting the answer, and returning the answer to the user, Since the system responds to questions, the probability that the user obtains an answer with desired contents increases.

In this specification, the knowledge management / retrieval system particularly in the medical field has been described. However, it goes without saying that the knowledge management / search system is not limited to the medical field, but can be applied to various fields.

Further, the knowledge management / retrieval system operated on the network by making the servers independent has been described. However, one server computer having a high processing capacity may be used to share the functions of a plurality of servers described in the present embodiment. These are determined in consideration of the processing capacity of the server, the operation mode of the system, and the like.

[0063]

According to the present invention, it is possible to provide a knowledge management / retrieval system which realizes both improvement of retrieval accuracy and efficient use of information, and which is easy for the user to use.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram illustrating an overall system for operating a knowledge management / search system according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram illustrating the concept of a knowledge management / search system according to the embodiment of this invention.

[Description of Signs] 1 Telecommunication equipment 2 User terminal 3 Communication server 4 E-mail server 5 Knowledge management / search system operation server 6 Database server 7 Operator terminal 10 Information correction engine 20 Categorize filter 30 Search engine 30a Search engine 30b Concept Dictionary Database 40 Answer Content Accuracy Improvement Engine 50 Expression Conversion Engine 60 Interest Recognition Engine

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G06F 15/403 340A

Claims (6)

[Claims] 1. A method for analyzing the contents of an e-mail about a question or question sent from a user, searching a knowledge database based on the analysis result, extracting an answer, and returning the answer to the user,・ In a knowledge management and retrieval system that responds to questions, the Japanese document data of the user's e-mail is compared with the standard document structure model to determine whether the Japanese document data is missing or incorrect. If it is determined that the Japanese document data in the mail is missing or incorrect, an e-mail is automatically created and sent to the user requesting that the Japanese text data be replaced or corrected. A control means for performing an information correction procedure to supplement and correct using the Japanese sentence data of the e-mail returned from to create accurate Japanese document data, A knowledge management and retrieval system characterized by comprising: 2. The knowledge management / retrieval system according to claim 1, wherein the information correction step counts the number of replies to an e-mail to the user, and automatically executes the replies if the number of replies reaches a predetermined number. Knowledge management characterized by including a procedure to create and reply to an e-mail requesting the completion or correction of errors in Japanese sentence data by manual input by an operator instead of manually creating and replying to e-mail・ Search system. 3. Analyzing the contents of an e-mail about a question or question sent from a user, searching a knowledge database based on the analysis result, extracting an answer, and returning the answer to the user,・ In a knowledge management and search system that responds to questions, search and extract keywords related to specific information from Japanese document data of e-mails from users, and record multiple knowledge databases classified for each information category. Access the database group, compare the category classification keywords assigned to each of the multiple knowledge databases with the previously searched and extracted keywords, and learn the knowledge database that matches the information category of the Japanese document data of the e-mail from the user A control means for performing a categorizing procedure selected from a group of databases, A knowledge management and retrieval system characterized by comprising: 4. The knowledge management and retrieval system according to claim 3, wherein the categorizing step retrieves and extracts keywords relating to specific information from knowledge data extracted from a knowledge database as a result of answer search, and classifies the keywords for each information category. Access the knowledge database group where multiple knowledge databases are recorded, compare the category classification keywords assigned to each of the multiple knowledge databases with the keywords retrieved and extracted earlier, and send a Japanese document of e-mail from the user Knowledge management / search including a procedure of selecting a knowledge database that matches the information category of data from a group of knowledge databases, and updating the database by adding the knowledge data previously searched and extracted to the selected knowledge database. system. 5. Analyzing the contents of an e-mail about a question or question sent from a user, searching a knowledge database based on the analysis result, extracting an answer, and returning the answer to the user,・ In a knowledge management and search system that responds to questions, the syntactic structure of the knowledge data extracted from the database as a result of the search is compared with various sample data, and the matching knowledge data is selected and extracted. A knowledge management / search system comprising control means for performing an improvement procedure. 6. The knowledge management / search system according to claim 5, wherein the answer content accuracy improving step compares the knowledge data extracted from the database as a result of the search with sentence sample data of various forms. Checks whether there is a problem in the syntax structure, checks whether the subject matter is reverse, checks patterns in the syntax structure, analyzes a plurality or all of these checks, and selects and extracts knowledge data Knowledge management and search system characterized by the following.