JP7311899B2 - Question answering system, question receiving and answering system, primary answer system, and question answering method using them - Google Patents

Description

TECHNICAL FIELD The present invention relates to a question answering system that accepts questions input from specified or unspecified end-user terminal devices and provides answers to the questions, and more particularly to a question answering system using artificial intelligence. , a question receiving and answering system and a primary answering system provided with artificial intelligence, which cooperate to constitute the question answering system, and a question answering method using them.

In recent years, in order to provide the optimal answer to the question, as a means of assisting specific end-users, such as telephone operators, in answering questions from customers, it has become possible to search past customer service histories and quickly provide more appropriate answers. Answer assistance system using artificial intelligence that supports answering questions from unspecified end users (questioners), various responses that automatically respond using artificial intelligence without human intervention system is being developed. Furthermore, natural language processing technology that searches for answers to questions by analyzing the syntax and meaning of questions entered by questioners, and research on extracting optimal answers using artificial intelligence are also underway.

In the following description of this specification, an unspecified or specific end user who asks a question is called a "questioner", and a person who introduces and uses the question answering system is called a "system user". A person who provides a question answering system to users is called a "system provider".
In addition, the terminal device used by the end user for asking questions is called the "question terminal", and the system user's system, which is the platform for the question reception page that is passed through when accessing the question answering system from the question terminal, is the "user system". ”. Furthermore, the system administrator who acquires, analyzes, and maintains the log information of the question-answering system used by the company as a system user is referred to as the "user's system administrator", and as the system provider, the operation and maintenance of the entire system. A person who manages maintenance, system improvement, etc. in general is simply referred to as a "system administrator."

For example, Patent Literature 1 discloses searching for answer candidates for a question, calculating a score indicating the degree to which each answer candidate is a correct answer to the question for each combination of a question and an answer candidate, and adding the score to the calculated score. A question answering system is disclosed that is configured to output an answer candidate that is most likely to be a correct answer to a question based on the above.
In addition, Patent Document 2 discloses a request content identification system that analyzes an input sentence describing a request from a user in a natural language and determines whether or not the input request content corresponds to a previously defined request content. It is

Furthermore, Patent Literature 3 discloses an information providing system including an information providing server and a suggestion server. When the user inputs a question, the suggestion server displays question candidates that may be input to assist the user in inputting the question. Further, the system of Patent Document 3 is configured so that maintenance of the information providing system can be performed by a system management terminal.

JP 2013-171550 A Patent No. 5099367 JP 2016-99967 A

However, in the systems of Patent Documents 1 and 2, it is necessary to collect a large amount of knowledge data and carry out machine learning in order to improve the accuracy of answering questions. In addition, advanced knowledge of artificial intelligence was required for operation management after the introduction of the question answering system.
Therefore, when introducing a question-answering system, a lot of time, effort, and enormous cost were required for collecting a huge amount of knowledge data and machine learning before starting operation. Furthermore, even after the system starts operating, machine learning for updating the knowledge of artificial intelligence and maintenance of the system require a lot of time and cost. It was necessary to secure human resources with specialized knowledge about intelligence.

Patent document 3, in view of problems in terms of operation management after system operation, in the maintenance mode, the system administrator uses the system management terminal to receive the response data log etc. from the response content determination means, It allows you to simulate input and verify the validity of knowledge data. In addition, the invention of this patent document 3 allows even if the system administrator does not have highly specialized knowledge about artificial intelligence, to change the expected question character string and response character string to artificial intelligence in the maintenance mode. It is intended to support

However, inputting a hypothetical question sentence and changing a response sentence to it change the knowledge data of artificial intelligence itself. Therefore, even in the system of Patent Document 3, although advanced knowledge is not required for collecting and checking log data and simulating input, system administrators are required to have a certain degree of specialized skills related to artificial intelligence. (See the latter part of (0048) of Patent Document 3). In addition, changing the knowledge data of artificial intelligence requires re-learning to reconstruct the neural network, which requires a lot of time and effort in addition to specialized knowledge of artificial intelligence. Therefore, in conventional question answering systems using artificial intelligence, system users who do not have specialized knowledge cannot directly participate in improving and gradually growing the question answering system introduced in their own company. rice field.
In addition, conventional question-answering systems using artificial intelligence do not have a simple mechanism to flexibly change specific answers to questions while actually operating the system. It was difficult to flexibly evolve to meet the needs of

Under the above circumstances, even if a small and medium-sized company with limited financial resources wishes to introduce a question answering system, the conventional question answering system requires a huge amount of cost to introduce, and it is difficult to secure professional human resources for system operation management. The hurdles to introduce the system were extremely high, as the need for such a system became a major barrier. Moreover, even if this barrier to introduction could be overcome, it would not have been possible for the users of the question answering system to directly improve the system and grow the system in line with changes in their company's needs. Furthermore, conventional question answering systems could not respond to questions in multiple languages simply by adding a simple configuration.

The present invention has been made in view of the problems of the conventional question answering system as described above and social needs, and can reduce the system introduction cost and easily change the final answer after the system starts operating. A question answering system using artificial intelligence that can improve and evolve the question answering system, a question reception and answer system that constitutes the same question answering system, a primary answer system, and a question answering method using them intended to provide
Another object of the present invention is to provide a simple configuration capable of accepting questions in a variety of languages with a simple configuration and providing a final answer to the question terminal in a predetermined answer language according to the language of the question. It is an object of the present invention to provide a question answering system, a question receiving and answering system constituting the question answering system, a primary answer system, and a question answering method using them.
Furthermore, other objects of the present invention are a question answering system using artificial intelligence that can be shared by a plurality of users and that can reduce introduction costs, a question receiving and answering system that constitutes the same question answering system, and a primary answer. It aims at providing the system and the question answering method using them.

In the present invention, a question answering system is configured by separating a question receiving and answering system for receiving questions and returning final answers to the questions to a question terminal and a primary answering system equipped with artificial intelligence, and the primary answering system is A primary answer to a question is provided as an identifier in response to a request from the question accepting and answering system, and the question accepting and answering system associates a final answer candidate with the identifier of the primary answer received from the primary answer system. The final answer candidate is read out based on the primary answer from the primary answer system, and the final answer is returned to the question terminal device.

A question answering system according to a first aspect of the present invention includes a question accepting and answering system using a computer that accepts questions from question terminals and returns final answers to the questions to the question terminals, and artificial intelligence. A question answering system separately provided with a primary answer system for replying a primary answer to a question in response to a request,
The primary response system is
receiving a reply request for the primary answer to the question from the question accepting and answering system, inputting the question received from the question accepting and answering system to the artificial intelligence, and providing the first answer to the question from the artificial intelligence interface means for transmitting the next answer to the question acceptance and answer system;
By machine-learning a plurality of assumed questions and a predetermined identifier as a primary answer corresponding to the assumed question, in response to a request from the question reception and answer system, the first answer to the question from the question reception and answer system and an artificial intelligence that outputs the primary response as the identifier,
The question reception and answer system is
receiving means for receiving questions from the question terminal;
A primary response inquiry that transmits the question received by the question receiving means to the primary response system, requests the primary response, and receives the primary response from the primary response system. means and
a final answer storage unit that associates 1:1 with the identifier machine-learned by the artificial intelligence and stores a final answer candidate corresponding to the identifier;
based on the primary response received from the primary response system, the final response candidate corresponding to the identifier of the primary response is read from the final response storage unit, and the read final response candidate is stored and a final reply means for returning a final reply to the question terminal based on the result.

Here, a general-purpose artificial intelligence system provided for the purpose of commercial services can be used as the primary response system that provides the primary response. It can be connected via a communication network. By separating the question answering system into the question receiving and answering system and the primary answering system in this way, the question receiving and answering system is separated from the artificial intelligence, so the machine of the artificial intelligence of the primary answering system Learning can be designed separately from the final answer candidates of the question receiving and answering system. Also, apart from machine learning of artificial intelligence, it is possible to independently modify, change, and add information separately from the hit rate of answers to questions.

A question accepting and answering system according to a first aspect of the present invention includes a question accepting and answering system using a computer that accepts questions from a question terminal, artificial intelligence, a plurality of assumed questions, and a question corresponding to the assumed questions. A question answering system in which a predetermined identifier is machine-learned as a primary answer, and a primary answer system that outputs a primary answer to a question from the question receiving and answering system as the identifier is provided separately. A question reception and answer system,
question receiving means for receiving questions from the question terminal;
A primary response inquiry that transmits the question received by the question receiving means to the primary response system, requests the primary response, and receives the primary response from the primary response system. means and
a final answer storage unit that associates 1:1 with the identifier machine-learned in the primary answer system and stores a final answer candidate corresponding to the identifier;
Based on the primary response received from the primary response system, the final response candidate corresponding to the identifier of the primary response is read out from the final response storage unit, and the read out final response candidate is stored. final answer means for returning a final answer to the question terminal based on
characterized by comprising
The invention according to this aspect is a sub-combination invention that constitutes a question receiving and answering system combining the question receiving and answering system and the primary answering system described above.

Further, in the question answering system or the question receiving and answering system according to another aspect of the present invention, the artificial intelligence of the primary answering system has a plurality of the above questions in which one or more assumed questions having a common attribute are grouped. Machine learning is performed so that a number of the assumed questions constituting a question group and one identifier corresponding to the question group on a 1:1 basis is used as a primary answer to the assumed question included in each question group. characterized by

In the above-mentioned question answering system or question receiving and answering system, the question receiving means can be further configured to acquire unique attribute information of the question accompanying the question from the question terminal, and the final answer storage unit can , if necessary, it can be configured to store different final answer candidates corresponding to the unique attribute information.
The unique attribute information includes identification information of the user system, language information indicating the type of language of the question, and the like. The identification information of the user system can be obtained based on communication route information when receiving a question or information added from the user system on the communication route. Even if the language information of the question is configured to be specified at the time of question input, it can be obtained by well-known techniques such as language specification information such as the OS of the question terminal or the application for question answering.

A question answering system or a question receiving and answering system according to another aspect of the present invention acquires identification information of a user system through which the question has passed, based on the acquired unique attribute information,
The final answer storage unit stores, as each final answer candidate, a basic answer that is the base of the final answer stored in association with each identifier, or, if necessary, each identifier and identification information of the user system. each of the basic answers stored in association with each other and unique answers having different contents are stored,
The final answer means reads out the basic answer or the unique answer from the final answer storage unit based on the identifier of the primary answer and the identification information of the user system corresponding to the question. The basic answer or the unique answer is transmitted to the question terminal as a final answer.
As described above, the identification information of the user system is obtained as unique attribute information from the communication route information through which the question arrived at the question reception and answer system or from the information given in the communication process. Is possible.

Further, in the question answering system or the question receiving and answering system according to another aspect of the present invention, the artificial intelligence of the primary answering system has a plurality of hypothetical questions having a common attribute consisting of different languages in one question group. All the hypothetical questions composed of multiple question groups including Machine learning is performed to output the primary answers with the same identifier regardless of type, and when a question is received from the question acceptance and answer system, the question is responded to as the primary answer to the question. and outputting the identifier to the interface means.
The artificial intelligence of the primary answer system is machine-learned to output the same primary answer for multiple hypothetical questions in different languages with common attributes. The same primary answer is output for the question. Since the final answer candidate is stored in association with the primary answer, if the primary answer is the same, the same final answer will be sent to the question terminal even if the language of the question is different. Become. The answer language of the final answer is the same regardless of the language of the question. It may be configured to send back the final answer in the same answer language as the question language.
Here, "a plurality of hypothetical questions having common attributes in different languages" includes hypothetical questions with substantially the same meaning in different languages.

In a question answering system or a question accepting and answering system according to another aspect of the present invention, the question accepting and answering system acquires question language information included in the unique attribute information of the received question by the question accepting means, The final answer storage unit stores the final answer candidates in a plurality of languages having substantially the same content in the answer language determined based on the question language information in association with the identifier and the answer language type. cage,
The final answer means acquires the information of the answer language corresponding to the question based on the acquired question language information, and based on the acquired identifier of the primary answer and the information of the answer language, The final answer candidate composed of the answer language corresponding to the question is read, and the final answer in the answer language corresponding to the question language is returned as the final answer corresponding to the question.
With this configuration, when the language of the question is a foreign language, it is possible to set the answer language to English or another language, or to reply the final answer in the same answer language as the question language.

In a question answering system or a question accepting and answering system according to another aspect of the present invention, the primary answering system extracts a primary answer to the question and outputs the answer confidence, and the question accepting and answering system reads from the final answer storage unit the basic answer or the unique answer stored in association with the primary answer according to the answer certainty received from the primary answer system, and responds to the question A final answer is transmitted to the question terminal.

In addition, the question answering system or the question receiving and answering system according to another aspect of the present invention further includes the following: An operation history storage unit that stores various information up to the final answer, and a user system that browses the various information stored in the operation history storage unit and stores the information in the final answer storage unit according to authority. and maintenance means capable of performing maintenance including correction, addition, and change of the final answer candidate or the unique answer.
This configuration makes it possible to easily maintain the primary answer system and the question answering system. In particular, maintenance of the question answering system does not require knowledge of artificial intelligence, and the system can be easily improved by changing the basic answers and unique answers that are final answer candidates.
Further, the content of the final answer candidates, the basic answers, or the unique answers stored in the final answer storage unit is restricted to the assumed questions of the primary answer system by the maintenance means according to authority. It can also be configured so that updates such as additions, corrections, changes, and deletions can be made without any modification.

The operation history storage unit stores the number of questions and answers for the entire system and for each system user, the average answer confidence, the number and ratio of answers whose answer confidence is below the reference value, the number of questions and answers for each identifier of the primary response and the average Information such as answer confidence, list of questions and answers with low answer confidence can be included. These pieces of information can be browsed by maintenance means, and these pieces of information make it possible to easily grasp the operating conditions of the system, such as the conditions of use of the system and the accuracy of responses. In addition, since the question receiving and answering system and the final answer storage unit are separated from the primary answer system, even if there is no expert knowledge about artificial intelligence, the final answers stored in the final answer storage unit can be obtained by maintenance means. and modify (in this specification, unless otherwise specified to the contrary, the term "modification" includes not only partial correction or change, but also addition can be used as Therefore, not only system administrators but also system users who do not have advanced knowledge of artificial intelligence can modify, change, or add final answers such as basic answers and unique answers within the given authority. It became possible.

The maintenance means can also be configured to add a function to modify, change, or add assumed questions, common attributes, question groups, or primary answers, and perform machine learning again on the primary answer system. With such an operation history storage unit and maintenance means, it is possible to output a list of extracted records of questions and answers with low answer certainty, and to make assumptions with low answer certainty according to analysis data from aggregation and statistical data. It is possible to integrate and abolish questions and question groups, and to improve and evolve the question answering system to provide more appropriate answers.

A first response system according to a first aspect of the present invention includes a question reception and answer system using a computer that receives questions from a question terminal and returns a final answer to the question to the question terminal, and a plurality of assumed questions. and a predetermined identifier as a primary answer corresponding to the assumed question by machine learning to output the primary answer to the question from the question receiving and answering system as the identifier. The primary answer system in the question answering system separately comprising
receiving a question from the question receiving and answering system and a request for the primary answer, inputting the question to the artificial intelligence, and sending a primary answer to the question from the artificial intelligence to the question receiving and answering system; interface means for transmitting through
a plurality of hypothetical questions in different languages, and a plurality of hypothetical questions, each of which has a common attribute regardless of language, as a plurality of question groups, each of which is 1: Each identifier corresponding to 1 is machine-learned as a common primary answer to the plurality of assumed questions belonging to the individual question groups, and when a question is received from the question acceptance and answer system, the question is answered. an artificial intelligence that outputs the primary response as the identifier to the interface means.

The present invention relates to a primary answering system that constitutes the question receiving and answering system described above, and is a sub-combination invention of the question answering system.
According to this configuration, the primary answer system is machine-learned so as to output one common primary answer (identifier) for assumed questions having common attributes even if the question languages are different. Therefore, the same identification number is output as the primary answer for questions having common attributes even if the languages of the questions are different. Therefore, even if the question language is different, the same final answer can be returned to the question terminal. Since it is relatively difficult to create an accurate question text in a language other than the native language, it is possible to ask an accurate question and obtain an accurate answer if the question can be asked in the native language. On the other hand, since the document reading ability is generally higher than the document writing ability, there is a high possibility that the respondent can understand the language even if it is not their mother tongue.
Further, by predetermining the answer language for the final answer according to the language of the question and storing the final answer candidate for each answer language, the final answer can be made in the predetermined answer language according to the language information of the question. It is also possible to configure so that an answer is sent back.

A question answering method for a question answering system according to a first aspect of the present invention comprises a question accepting and answering system for accepting a question from a question terminal by a computer and replying an answer to the question to the question terminal, and artificial intelligence. machine-learning a plurality of assumed questions and predetermined identifiers that serve as answers to the plurality of assumed questions as primary answers, and returning identifiers as primary answers according to the input questions. A question answering method in a question answering system comprising:
(a) in the question acceptance and answer system, a step of accepting a question from a question terminal and transmitting the question to a primary answer system;
(b) In the primary response system, the question is received from the question acceptance and response system, the question is input to the artificial intelligence, and the primary response to the input question is output as the identifier. and sending the identifier of the primary answer output from the artificial intelligence to the question receiving and answering system;
(c) In the question receiving and answering system, when the identifier is received as the primary answer from the primary answering system, it is linked to the identifier machine-learned by the artificial intelligence and the final answer candidate a step of reading the final answer candidate corresponding to the received identifier from the final answer storage unit storing the final answer candidate, and replying the read final answer candidate to the question terminal as a final answer;
characterized by comprising

A question answering method of a question accepting and answering system according to a first aspect of the present invention comprises a question accepting and answering system for accepting a question from a question terminal by a computer and replying an answer to the question to the question terminal, and artificial intelligence. Machine learning is performed on a plurality of assumed questions and predetermined identifiers that are answers to the plurality of assumed questions as primary answers, and identifiers are returned as primary answers according to the input questions. A question answering method of the question receiving and answering system in a question answering system comprising:
(a1) receiving a question from a question terminal in the question receiving and answering system and transmitting the question to the primary answering system;
(b1) When the identifier is received as the primary response from the primary response system, the final response is linked to the identifier machine-learned by the artificial intelligence and stored as a final response candidate. a step of reading the final answer candidate corresponding to the received identifier from a storage unit and replying the read final answer candidate as a final answer to the question terminal;
characterized by comprising

In the question answering method of the question answering system or the question accepting and answering system according to another aspect of the present invention, the step (a) or (a1) is
(a-1) receiving a question from a question terminal;
(a-2) acquiring and storing unique attribute information of the question accompanying the question;
(a-3) sending the question to the primary response system;
with
The step (c) or (b1) is
(c-1) receiving the identifier as the primary response from the primary response system;
(c-2) obtaining identification information of the user system from the unique attribute information stored in step (a-1);
(c-3) Identifying the identifier of the primary response and the user system from the response storage unit that stores the final response candidates corresponding to the identifier, divided into basic responses and user-specific unique responses. retrieving the corresponding basic answer or the unique answer based on the information;
(c-4) returning the read basic answer or unique answer to the question terminal;
characterized by comprising

A multilingual question answering method for a question answering system according to a first aspect of the present invention includes a question accepting/answering system for accepting a question from a question terminal by a computer and replying an answer to the question to the question terminal; All the hypothetical questions composed of a plurality of question groups including a plurality of hypothetical questions having common attributes in different languages in one question group, and different languages included in the same question group Machine learning is performed so that the primary answers to the plurality of assumed questions are output with the same identifier regardless of the language type for any of the assumed questions, and the input question A multilingual question response method in a question answering system comprising a primary response system that returns the identifier as a primary response in response to
(a) In the question receiving and answering system, receiving a question from a question terminal, acquiring and storing language information of the question from unique attribute information of the question accompanying the question, storing the question and the primary if an answering system requires the language information of the question, sending the question to the primary answer system together with the language information of the question;
(b) In the primary response system, the question is received from the question acceptance and response system, the question is input to the artificial intelligence, and the primary response to the input question is output as the identifier. and sending the identifier of the primary answer output from the artificial intelligence to the question receiving and answering system;
(c) In the question receiving and answering system, when the identifier is received as the primary answer from the primary answering system, it is linked to the identifier machine-learned by the artificial intelligence and the final answer candidate a step of reading the final answer candidate corresponding to the received identifier from the answer storage unit storing the identifier, and replying the read final answer candidate to the question terminal as a final answer;
characterized by comprising
Here, "a plurality of hypothetical questions having common attributes in different languages" includes hypothetical questions with substantially the same meaning in different languages.

A method for accepting and responding to multilingual questions in a question accepting and answering system according to a first aspect of the present invention comprises: a question accepting and answering system for accepting questions from question terminals by a computer and replying answers to the questions to the question terminals; All the hypothetical questions that are equipped with artificial intelligence and are composed of a plurality of question groups including a plurality of hypothetical questions that have a common attribute consisting of different languages in one question group, and the different languages included in the same question group. The primary answers to the plurality of assumed questions consisting of are input by machine learning so that the primary answers with the same identifier are output for all the assumed questions regardless of the type of language A question acceptance and response method of the question acceptance and answer system in a question answering system comprising a primary answer system that returns an identifier as a primary answer in response to a question,
(a1) In the question receiving and answering system, receiving a question from the question terminal, acquiring and storing language information of the question from unique attribute information of the question accompanying the question, and storing the question and the first if the next answer system needs the language information of the question, sending the language information of the question together to the first answer system;
(b1) When the identifier is received as the primary response from the primary response system, the final response is linked to the identifier machine-learned by the artificial intelligence and stored as a final response candidate. a step of reading the final answer candidate corresponding to the received identifier from a storage unit and replying the read final answer candidate as a final answer to the question terminal;
characterized by comprising
The present invention is a sub-combination invention of the invention according to claim 16. Here, "a plurality of hypothetical questions having common attributes in different languages" includes hypothetical questions with substantially the same meaning in different languages.

In the multilingual question reception and response method of the question answering system or the question reception and answer system according to another aspect of the present invention, the step (c) or (b1) includes:
(c-1) receiving the identifier as the primary response from the primary response system;
(c-2) a step of obtaining answer language information determined according to answer language determination rules based on the language information stored in step (a) or (a1);
(c-3) The final answer candidate stored in the final answer storage unit is linked to the machine-learned identifier and the answer language information, and the final answer candidate stored in the final answer storage unit is selected from the answer language specified by the answer language information. a step of reading out the final answer candidate corresponding to the identifier received as an answer and the answer language information obtained;
(c-4) a step of replying the read out final answer candidate in the answer language as a final answer to the question terminal;
characterized by comprising

An answering method of a primary answering system according to a first aspect of the present invention includes a question accepting and answering system for accepting a question from a question terminal and replying an answer to the question terminal, and responding to a request from the question accepting and answering system. and a primary answer system provided with artificial intelligence that outputs a primary answer to the question as an identifier, and a method of answering the primary answer system in a question answering system using a primary answer system,
(a) One question group includes a plurality of hypothetical questions having a common attribute regardless of language, and a large number of hypothetical questions that constitute the plurality of question groups, and each of the plurality of question groups correspond to each other on a 1:1 basis. causing the artificial intelligence to machine-learn the identifier as a primary answer to each of the hypothetical questions;
(b) inputting the question received from the question receiving and answering system into the artificial intelligence and outputting the identifier as a primary answer from the artificial intelligence;
(c) transmitting the identifier output as the primary response to the question acceptance and response system as the primary response to the question;
In step (a), the plurality of hypothetical questions having a common attribute included in each of the question groups includes hypothetical questions that have substantially the same meaning but differ in language. preferably.
The present invention is a sub-combination invention of the invention according to claim 16.

In the question answering system of the present invention, after separating the question reception and answer system and the primary answer system having artificial intelligence, the final answer candidate to the question is sent to the question reception and answer system, and the first answer system of the primary answer system. It is configured so that the final answer candidates of the question receiving and answering system are read out according to the primary answer from the primary answering system, and the final answer is sent back. As a result, the contents of the final answer candidate can be freely corrected and changed regardless of artificial intelligence. That is, even without expert knowledge of artificial intelligence, the final answer can be freely changed within the scope of the predetermined authority. Even if there is, by changing the final answer to the question so that it is the final answer according to the question, it is possible to improve and maintain such that the answer confidence is substantially increased without machine learning .

In addition, in order to improve the accuracy of responses to questions, it is desirable to use machine learning for as many possible questions as possible from the beginning of system introduction. get higher On the other hand, in the present invention, since the final answer candidate is stored in a separate question reception and answer system separate from the primary answer system, the content of the answer is expanded to a wider range than the content of the direct answer to the assumed question. It is possible to store final answer candidates. As a result, to the extent that the primary answer can obtain a certain degree of answer confidence, answers to hypothetical questions that have not been machine-learned by artificial intelligence are also included in the answer content of the final answer candidate and stored, so that it is possible to respond to the question. It is possible to increase the satisfaction of the answer. Therefore, in the present invention, it is possible to improve the answer contents of the final answer candidates while reducing the number of assumed questions and answers to be machine-learned by the primary answer system. In other words, it is possible to reduce the number of assumed questions to be machine-learned by artificial intelligence, lowering the introduction cost of the question answering system at the start of operation, and reducing the load of machine learning of artificial intelligence at the start of operation. It becomes possible to

Further, in the question answering system according to another aspect of the present invention, after separating the question receiving and answering system and the primary answering system, by storing together different final answer candidates for each system user, a plurality of system users can jointly use one question answering system. This makes it possible for individual system users to reduce the cost of introducing the system.

In addition, in the question answering system according to another aspect of the present invention, after separating the question receiving and answering system and the primary answering system, even if there are a plurality of assumed questions in different languages, only one assumed question having a common attribute is used. It is possible to receive questions in different languages with a simple configuration by performing machine learning so as to obtain two identical primary answers and storing final answer candidates in different answer languages in the question accepting and answering system. The answer language can be configured to answer in one language, or can be configured to answer in a plurality of predetermined answer languages according to the language of the question.

FIG. 1 is a conceptual diagram for explaining how the question answering system according to the first embodiment of the present invention is used; FIG. 4 is a schematic diagram showing a configuration example of another embodiment (second embodiment) of the question answering system of the present invention; 3(a) is a functional block diagram illustrating an embodiment of the question receiving and answering system of the present invention, and (b) is a functional block illustrating a configuration example of final answering means shown in FIG. 3(a). 4 is a functional block diagram showing a configuration example of maintenance means shown in FIG. 3; FIG. 4 is a flow chart illustrating the flow of responses to questions according to the present invention. FIG. 4 is a diagram showing specific examples of common attributes and corresponding relationships between primary responses and common attributes for question groups Q1 to QN that are learned by a primary answer system such as AI used in the present invention. FIG. 4 is a diagram for explaining by exemplifying question groups and primary answers according to the present invention; FIG. 10 is a diagram for explaining by exemplifying other question groups and primary answers according to the present invention; FIG. 11 is a diagram for illustrating and explaining still another question group and primary answers; FIG. 4 is a diagram illustrating the relationship between a primary reply identifier and a final reply stored in a final reply storage unit according to the present invention; FIG. 11 is a diagram exemplifying the contents of the basic answer of the final answer corresponding to the identifier "NNN" shown in FIG. 10; 4 is a flowchart illustrating basic procedures for initial setting; Flowchart exemplifying the procedure for maintenance of the final reply in the final reply storage unit by the trader 5 is a flow chart illustrating maintenance procedures for re-learning the question answering system and reviewing basic answers regularly or irregularly; FIG. 10 is a diagram showing specific examples of common attributes of assumed questions of each question group and corresponding relationships of primary answers corresponding to the question groups with respect to question groups Q1 to QN in an administrative agency (for example, a city hall or a ward office); The figure which shows the relationship between the specific example of the assumption question of the question group Q1 of multiple languages, and its primary answer. The figure which shows the relationship between the specific example of the assumption question of the question group Q2 of multiple languages, and its primary answer. FIG. 10 is a diagram showing a relationship between a concrete example of assumed questions of a question group QN in multiple languages and primary answers thereof; FIG. 10 is a diagram illustrating the relationship between primary response identifiers, language identifiers, and final responses in multiple languages; 4 is a flow chart illustrating a processing procedure for multiple languages; FIG. 10 is a diagram illustrating specific examples of common attributes for question groups Q1 to QN for explaining entities and corresponding relationships with primary answers; FIG. 10 is a diagram showing the relationship between assumed questions and primary answers of the question group Q1 for explaining entities; FIG. 10 is a diagram showing the relationship between assumed questions and primary answers of question group Q2 for explaining entities. FIG. 10 is a diagram showing the relationship between assumed questions of question group Q3 and primary answers for explaining entities. FIG. 10 is a diagram showing the relationship between assumed questions and primary answers of question group QN-1 for explaining entities. FIG. 10 is a diagram showing the relationship between assumed questions and primary answers of the question group QN for explaining entities; FIG. 10 is a diagram illustrating the relationship between primary reply identifiers and final replies; Diagram explaining the correspondence of the primary answer corresponding to the intent question group

A question answering system, a question receiving and answering system, a primary answering system and a question answering method according to the present invention will be described below with reference to the drawings. FIG. 1 is a diagram for schematically explaining a usage form and an overall schematic configuration of a question answering system according to a first embodiment of the present invention. In FIG. 1, a question sent from a question receiving/answering system 10 and a primary answering system 30, which constitute a question answering system 25 according to the present invention, and a question terminal 90 are received by the question answering system 25 of the present invention, and the question is It illustrates part of various devices and communication networks that work together until the final response is returned to the terminal 90 .

Questions from the question terminal 90 are sent via a question acceptance page on a system (hereinafter referred to as "user system") 40 that serves as a platform for users (traders, etc.) connected via a communication network 91. It is transmitted to the question receiving and answering system 10 . If the question terminal 90 is a terminal device such as a desktop PC or a smart phone connected to the user system 40 , questions are sent directly from the user system 40 to the question reception and answer system 10 . Here, the communication network 91 includes facilities such as a wireless communication network and relay stations (not shown) and a plurality of networks (including the LAN of the system provider). 1 and 2 show the form in which the user system 40 is connected to an external network, it may be connected to the local area network of the provider who provides the question answering system. .

1 to 14, traders A to N are exemplified as users of the question answering system. However, the users of the question answering system of the present invention are not limited to so-called commercial or industrial traders. That is, it is not limited to profit-making corporations and profit-making organizations, and any organization or individual including, for example, universities, hospitals, local governments, associations, etc., can be a user (system user) of the question answering system according to the present invention.

In FIG. 1, reference numeral 25 denotes a question answering system according to the first embodiment of the present invention, which includes a question receiving and answering system 10 and a primary answering system 30. As shown in FIG. The primary answer system 30 has a knowledge database such as artificial intelligence, and the final answer storage unit 18, which constitutes a part of the question reception/answer system, stores final answer candidates, which are answers to questions. The final answer storage unit 18 can be configured from a storage device such as a server or database. The final answer storage unit 18 may be a storage device inside the question receiving and answering system 10, may be connected via a LAN, or may be connected via an external communication network. In FIG. 1, the final answer storage unit 18 is indicated by a dashed line in the sense that it may exist as a storage device inside the question receiving and answering system 10 or as a storage device connected to a network. . FIG. 3 shows a functional block diagram as a configuration example showing the main functions of the question receiving and answering system 10. As shown in FIG.

The question receiving and answering system 10 is composed of hardware such as a server equipped with a CPU, ROM, RAM, etc., basic software and other various software, and includes question receiving means for receiving questions via a communication network 91; It is a computer system comprising a primary response inquiry means for inquiring a primary response to the primary response system 30, a final response means for returning a final response to the question terminal 90, a final response section, and the like.
The primary response system 30 includes interface means for receiving questions and replying responses from artificial intelligence, and artificial intelligence for outputting primary responses as identifiers by machine learning of assumed questions and responses. It is a computer system with

In the conventional question answering system, it was necessary to machine-learn a huge amount of hypothetical questions and answers in order to increase the satisfaction of the answers. As a result, a lot of time and effort have been required to collect a huge amount of assumed question data and answer data used for machine learning of artificial intelligence, and to perform machine learning on these data. In order to improve and update the accuracy of the question answering system, artificial intelligence machine learning was essential.
On the other hand, in the question answering system 25 of the present invention, the question receiving/answering system 10 and the primary answering system 30 equipped with artificial intelligence are clearly separated. and the final answer that actually replies. As a result, the primary answering system 30 equipped with artificial intelligence and the question receiving and answering system 10 that returns the final answer can be separated and improved and evolved separately. That is, the question answering system 25 according to the present invention can improve and evolve the system by machine learning artificial intelligence. By appropriately adding and changing the contents of the question answering system 25, it becomes possible to evolve and improve the question answering system 25 so that more appropriate answers can be given to questions.

In addition, in the present invention, the system can be constructed on the premise that the final answer candidate can be easily changed and improved in this way. can be reduced compared to the conventional method, and the final answer candidates can be enriched. As a result, the cost at the time of introduction and the machine learning load at the time of introduction can be reduced.

First, hypothetical questions that are machine-learned by the primary answer system 30 equipped with artificial intelligence and primary answers that are output from the primary answer system 30 as answers to the questions will be described.
For conventional artificial intelligence, machine learning (learning with supervision) is performed on a plurality of hypothetical questions and answers to the hypothetical questions. Using the example of a question-answering system at a city hall, in general, questions such as "Where should I go to pay my taxes?" machine learning to output an answer such as "The third window on the second floor is the tax-related window." However, in the primary answer system 30 of the present invention, machine learning is performed so that a specific identifier is output as the primary answer instead of the specific answer to the assumed question. Here, the form of the identifier may be of any type as long as it is a unique identifier that can identify the primary response, but it is desirable to have a short and simple structure as long as it can identify each primary response. In this specification, it is assumed that a simple three-digit numeric identifier such as "100" is output.

In other words, in the primary answer system 30, for a plurality of assumed questions having some common attribute (hereinafter simply referred to as "common attribute"), one same identifier is given as the primary answer for each assumed question. Machine learning to output. The size and range of common attributes can be set arbitrarily.

A set of hypothetical questions having one common attribute is referred to herein as a question group. That is, a plurality of assumed questions with a common meaning belong to one question group, and one common primary answer ( machine learning to output identifier). In this case, the number of primary answers (identifiers) is the same as the number of question groups, and the same primary answers (identifiers) are output for each assumed question belonging to the same question group. For example, in the example of the hypothetical questions about tax payments above, if the answers corresponding to the questions about "taxes" are grouped into one question group, all the hypothetical questions about taxes have one common identifier as the primary It will be machine-learned so that it will be output as an answer.

What is to be the common attribute can be determined according to the question group. For example, in the example of the hypothetical question about "Payment of taxes" above, "Where to pay taxes" is a common attribute. Here, it can be freely determined whether "tax refund" and "tax consultation" are included in the same question group or in separate question groups. For example, by broadening the scope of common attributes, all matters related to taxes can be treated as one answer. In this case, one question group is formed with "tax" as one common attribute, and primary answers to many assumed questions about taxes are all output as one identifier.
Upon receiving a question from the question receiving and answering system 10, the primary answering system 30 returns a predetermined identifier together with the answer certainty to the question receiving and answering system as a primary answer to the question by artificial intelligence.

In the question accepting and answering system 10, the final answer candidate is stored in the final answer storage unit 18 in association with the identifier (primary answer) machine-learned by the primary answer system. Each final answer candidate preferably includes at least all answers to multiple hypothetical questions with common attributes included in each question group, and only direct answers to machine-learned hypothetical questions Instead, it can also be configured to include answers to related questions as well. As a result, it is possible to reduce the system introduction period and the introduction cost while maintaining the satisfaction of the answers. Therefore, the final answer candidates in the present invention tend to be long because the final answer candidates include many answers such as answers to a plurality of hypothetical questions and other predicted answers. When the question receiving and answering system 10 receives a primary answer (identifier) from the primary answering system 30, in order to read the final answer candidate corresponding to the identifier, the final answer storage unit is linked to each identifier. 18.

The final answer storage unit 18 stores, as final answer candidates, at least basic answers serving as the basis for answers to all assumed questions belonging to each question group, linked to identifiers corresponding to the primary answers of each question group. ing. When only the basic answer is stored as the final answer candidate corresponding to each identifier, the basic answer is directly the final answer for that identifier. The contents of the basic answer preferably include answers to all assumed questions related to the question group corresponding to the linked identifier. More preferably, even if it is not included in the expected questions of the question group, it is desirable to include answers to expected questions with the same common attribute as the question group. In addition to basic answers, unique answers can be stored as final answer candidates in the final answer storage unit 18 . A unique answer is a final answer candidate that has the same basic framework as the basic answer but replaces the basic answer or adds new information to the basic answer. Therefore, when the question receiving and answering system 10 receives the primary answer (identifier) to the question from the primary answering system 30, the final answer candidate corresponding to the received identifier is selected from the final answer candidates in the final answer storage unit 18. , either the basic answer or the unique answer is read out and sent back to the question terminal 90 as the final answer to the question.

As will be described later using a specific example, the unique answer is used when the content of the answer to the question differs depending on the system user.
For example, based on the unique attribute information of the question transmitted from the question terminal 90, information identifying the system user (identification number, etc.) is acquired, and received from the final response storage unit 18 and the primary response system 30 Read out the primary response (identifier) and the "basic response or unique response" linked to the system user's identification number. As a result, even if the primary reply has the same identifier, it is possible to read out the unique reply corresponding to the identification number of the system user and return a different final reply for each system user.

Here, the “specific attribute information of the question” is information specific to each question received from the question terminal 90. For example, information about the question terminal 90, information other than the question input by the questioner, It can include various types of information such as communication information and communication route information from to the question answering system, information on the other party of the question (system user's identification number etc.), information on the language of the question and the language of the answer. Even if the question receiving and answering system 10 or the like automatically acquires these "question specific attribute information" from communication information, communication route information, etc., the questioner is requested to input the information as necessary. It may be configured. Further, the identification number of the system user may be added to the system user's identification number in the user system 40 when the question is transmitted via the user system 40 .

As a typical form of usage using a unique response, there is a case where it is desired to send back a final response that is different from the content of the basic response and to which business hours and locations unique to the system user are added. For example, there is a case where it is desired to inform a system user who is a real estate agent of the business hours and location specific to the system user (real estate agent) in response to a question about the real estate agent from the question terminal. In such a case, as a response corresponding to the primary response, in addition to the basic response, a unique response including the business hours and location of the system user (real estate agent) is stored in the final response storage unit. As a result, based on the primary response and the identification number of the system user, the unique response of the system user is read, and the business hours and location of the system user are returned to the question terminal 90 as the final response. will be

In addition, as will be explained later in detail, the artificial intelligence of the primary answer system is made to include hypothetical questions in different languages that have common attributes in the question group, and perform machine learning of hypothetical questions and primary answers. Thus, it is possible to accept questions in multiple types of question languages. The language of answers to questions does not necessarily have to be the same language as the language of the question, and the final answer may be returned in the standard language of the question answering system. Since the ability to read documents is relatively high compared to the ability to write documents, there are many people who can read and understand the answers even if they are not native speakers, so there is a high need to answer in the same language as the question language. do not have. Multilingualization of only the question language is useful in order to reduce the cost of creating answers and shorten the introduction period. If the introduction cost and introduction time are a little more generous, it is also useful to limit the answer language to languages such as Japanese, English, or Chinese that are likely to be used. It is also possible to configure the final answer in the same language as the question language. By linking the final answer in each answer language to each identifier of the primary answer and answer language information corresponding to the question language, and storing the final answer in each answer language in the final answer storage unit 18, It is possible to send back final answers in multiple answer languages.

As is clear from the above description, the present invention separates the primary response system 30 and the question reception and response system 10, artificial intelligence machine learning in the primary response system 30, and question reception and response system 10 It is configured so that the final reply can be created and updated independently of each other. In addition, the primary response system 30 and the question reception and response system 10 are connected by the concept of the common attribute of the assumed question and the identifier of the primary response. Therefore, the hypothetical question to be machine-learned, the primary answer (identifier), and the final answer can be matched, and the hypothetical question and the final answer candidate can be connected to each other. Therefore, it is possible to construct a final answer candidate with a flexible structure for an assumed question to be subjected to machine learning. system 25 can be performed independently.

As described above, since the primary response system 30 and the question reception and response system 10 can be separated and final response candidates can be flexibly constructed, the system introduction cost and load can be reduced according to the field of the question answering system to be applied. Various devices and efforts to reduce it are possible.
For example, by narrowing down to a specific field of a specialized field or a specific industry, prediction of questions and required answers become relatively clear. The final answer candidates of the reception/answer system can include answer contents corresponding to questions that have not been machine-learned by artificial intelligence. As a result, it is possible to improve the quality of the final answer while suppressing the number of questions assumed by machine learning and the number of primary answers. That is, since the load at the time of introduction can be greatly reduced, it is possible to reduce the introduction cost and the time until the start of operation.
Also, by narrowing down the range of target questions to questions in a specialized field, a specific industry, and a specific field, the number of expected questions can be narrowed down. By narrowing it down to a specific field of specialization or a specific industry, it becomes relatively easy to predict questions, so the number of types of common attributes can be reduced.

Furthermore, basically, the final answer candidates include all answers to the assumed question having common attributes, and furthermore, answers related to the assumed question. Therefore, for one question, a final answer including many frequently asked answers in the field related to the question is simultaneously provided for one question. Therefore, there is a high possibility that the final answer also includes the answer to the question that the questioner planned next. .

By configuring the primary response system 30 and the question reception and response system 10 separately in this way, the number of assumed questions and primary responses is reduced, the load of machine learning is reduced, and the construction and introduction costs of the question response system 25 are reduced. can be lowered. In addition, by narrowing down the range of questions that can be accepted, it is possible to greatly narrow down the number of expected questions for machine learning and the number of primary answers (identifiers) by narrowing down the number of common attributes while organizing the expected questions. can. For example, in order to increase the correct answer rate, conventionally, at least tens of thousands of assumed questions and answers are required. By narrowing it down to about 1000 hypothetical questions and primary answers and machine learning them, it became possible to start the operation of the question answering system.

In addition, by narrowing down the industry and business, it is possible to create highly accurate hypothetical questions that are standardized to correspond to the industry and business, and build a question-answering system. It is also possible to If a plurality of people in the same trade can jointly use a single question answering system, the introduction cost can be jointly borne, so that it is possible to further lower the barriers to system introduction.
As for the shared use of the single question answering system 25 by the same traders, as described above, the unique answers are stored in the final answer storage unit 18, and a different final answer can be returned for each user based on the unique attribute information of the question. By adopting such a configuration, it becomes possible to give individual final answers according to the individuality of each user, so that it is possible to realize a more desirable form of shared use.
The "common attribute", "assumed question", "primary answer", "basic answer", and "unique answer" explained above will be explained in more detail later using specific examples.

In the primary response system 30 for providing the primary response of the present invention, WATSON (WATSON is a registered trademark) of International Business Machines Corporation can be used as a general-purpose service of artificial intelligence. .

In the above description, a configuration example in which the primary answer system 30 is connected to the primary answer system 30 via the communication network 91 has been shown as the first embodiment of the question answering system.
However, as shown in FIG. 2 as an example, a question answering system 25' according to another embodiment of the present invention connects a question receiving and answering system 26 and a primary answering system 27 without an external communication network. , may be connected via an internal line or a dedicated communication line. In the configuration of FIG. 2, the only difference is that the question reception and answer system 26 and the primary answer system 27 are connected by an internal line (dedicated line, etc.). The basic configuration and functions are the same as those of the question receiving and answering system 10 and the primary answering system 30 of FIG. 1, respectively.

FIG. 1 shows an example in which a management terminal (system management terminal 50 ) of a system provider who operates and manages the question answering system 25 is connected to a communication network 91 . As is clear from the explanation that the communication network 91 also includes a LAN, the system management terminal 50 can also be configured to be connected to the question answering system 25 (25') via an internal communication system (LAN).

Next, a main configuration example of the question receiving and answering system 10 (26) will be described with reference to FIG. 3(a). FIG. 3(a) is a functional block diagram illustrating the configuration of the question receiving and answering system 10 (26) of the present invention. hardware, software for controlling the hardware, various programs, and the like.
The question receiving and answering system 10 (26) basically comprises a communication interface 11 having a transmission/reception function, receiving means 12 for receiving questions from the outside, a question terminal 90, and an operation history storage section storing information on questions. 13, primary answer inquiry means 14, and final answer means 15. It is also desirable to have maintenance means 16 .

The accepting means 12 of the question accepting/answering system 10 (26) accepts questions from the question terminal 90, and stores unique attribute information of the questions in, for example, the operation history storage unit 13 or the like. The primary response inquiry means 14 inquires the primary response system 30 (27) about the question received from the reception means 12, and receives the primary response and, if possible, the primary response system 30 (27). receives its answer confidence. Upon receiving the primary response and the response certainty from the primary response system 30 ( 27 ), the primary response inquiry means 14 outputs the information to the final response means 15 . The final response means 15, when the primary response received from the primary response means satisfies a predetermined condition (for example, when the response certainty is equal to or greater than a predetermined threshold value), the primary response and the unique attribute of the question The final answer candidate (basic answer or unique answer) corresponding to is read from the final answer storage unit 18 (28) and sent to the question terminal 90. The operation history storage unit 13 stores the operation history of each process from the question reception to the final answer, for example, the unique attribute of the question, the question content, the primary answer, the answer confidence, the final answer, etc., as necessary. . In the above and following explanations, an example is shown in which the response certainty is always received together with the primary response from the primary response system 30 (27). It is not mandatory to receive a degree. It is also possible to adopt a configuration in which other similar information is received in place of the answer certainty.

In the above example, one primary response (identifier) is received from the primary response system 30 (27). It is also possible to receive and select an appropriate unique primary response (identifier) in the question receiving and answering system 10 (26) to obtain a final response.

The maintenance unit 16 receives requests for collection, analysis, and maintenance of log information and various operating information from system administrators or system users, and executes processes such as information collection and analysis and maintenance. By the maintenance means 16, even a system user who does not have expert knowledge of artificial intelligence, for example, can store the final answer independently from the system administrator of the question answering system 25 (25') within a predetermined authority range. A part of the content of the final answer in the part 18 (28) can be independently managed, corrected, and changed.

FIG. 3(b) is a functional block diagram showing an example of the configuration of the final reply means 15 of FIG. 3(a).
When the primary response is received from the primary response system 30 (27), the primary response is transmitted from the primary response inquiry means 14 to the final response means 15. FIG. When the final answer control unit 35 of the final answer means 15 receives the primary answer, it acquires the unique attribute information of the question stored in the operation history storage unit 13 via the unique attribute information acquisition unit 36, It is transmitted to the reply control section 35 . The final response control unit 35 sends the primary response from the final response storage unit 18 (28) to the final response acquisition unit 37 based on the primary response identifier and unique attribute information (system user information, etc.). command to obtain a basic answer or a unique answer corresponding to The basic answer or unique answer acquired by the final answer acquisition unit 37 is sent to the final answer transmission unit 38, and under the control of the final answer control unit 35, it is sent to the question terminal 90 via the communication interface 11 as the final answer. be.

FIG. 4 is a functional block diagram showing a configuration example of the maintenance means 16 shown in FIG. 3(a). The maintenance means 16 comprises an initial setting means 21 , an analysis means 22 and a data update means 23 . Receiving operations from the system management terminal 50 via the initial setting means 21, inputting and setting multiple assumed questions and primary answers (identifiers) during system construction, and machine learning of the primary answer system 30 (27) Work such as data setting is performed, and final answer candidates (basic answers and, if necessary, unique answers) are stored in the final answer storage unit 18 (28).
The analysis means 22 analyzes statistical information such as questions, primary answers, final answers, and answer confidence stored in the operation history storage unit 13, for example, the number of questions and answers for each system user, questions and answers for each answer group, Create various statistical data such as the number of responses, average response confidence, and the number and ratio of responses whose response confidence is below the reference value. These statistics can only be monitored by authorized persons.

The data update means 23 receives access from the management terminal of the user system 40 and the system management terminal 50 and updates desired data.
From the system management terminal 50, it is possible to update the knowledge data of the primary answer system 30 (27) by machine learning, and to update the contents of the final answer candidates in the final answer storage unit 18 (28). can. For example, when the answer certainty of the primary answer to the question is low, a new hypothetical question and primary answer (identifier) are added and machine learning is performed by the artificial intelligence machine of the primary answer system 30 (27). be able to. In addition, it is possible to create a final answer candidate corresponding to a new identifier in the final answer storage unit of the question receiving and answering system, and to add, modify, and change existing basic answers and unique answers. The initialization means 21 can also integrate its functions into the data updating means 23 .

The system user can read out the access log information of the system user from the management terminal of the user system (40), view the operation history and analysis data, and within a predetermined permitted range, the final reply storage unit 18 The final answer data stored in (28) can be corrected and changed.
As a result, regardless of the presence or absence of specialized knowledge about artificial intelligence, system users (traders) themselves can change, modify, add, etc. the content of the final answer stored in the final answer storage unit 18 (28). That is, the system user (such as a trader) can use the management terminal of the user system 40 connected via the communication network within a predetermined and permitted range without requesting the system provider to make a final reply. The final answers (basic answers and unique answers) stored in the storage unit 18 can be corrected and updated.

With reference to FIG. 5, the basic flow of responses to questions from the question terminal 90 in the question answering system 25 (25') according to one embodiment of the present invention will be described. In the following explanation, an example of the processing procedure for providing a unique answer using the unique attribute information of the question will be explained. You are free to choose whether or not to provide an answer.
When a question q1 is received from the question terminal 90 (step S1), the reception means 12 stores the unique attribute information of the question, the content of the question, etc. in the operation history storage unit 13 (other storage units may be used) (step S2), the primary response inquiry means 14 inquires of the primary response system 30 (27) about the primary response to the question q1 (step S3).

The primary response system 30 (27) extracts the primary response (identifier of the final response: a1 to an in FIG. 5) to the question q1 and returns it to the primary response inquiry means . The primary response system 30 (27), for example, sends one or more primary responses (identifiers) in descending order of response certainty, 14. When the primary response inquiry means 14 receives the primary response from the primary response system 30 ( 27 ) (step S 4 ), it is sent to the final response means 15 . The final reply means 15 determines whether the primary reply (identifier) satisfies a predetermined condition for replying to the questioner. It is verified whether or not the answer certainty of the next answer exceeds a predetermined threshold (step S5).

If the first response satisfies the predetermined condition (step S5; Yes), the final response means 15 acquires the unique attribute information of the question and obtains the identification information of the system user (step S6). . The unique attribute information of the question is stored in the operation history storage unit 13 or the like. It is possible to identify the system user (trader) by checking whether the person has come. Next, based on the first response and the information (identification information) of the system user (vendor) obtained, basic answer candidates corresponding to the first response or unique answer candidates corresponding to the system user (vendor) are generated. acquire (step S7). If there are multiple primary answers received, the final answer candidates (basic answer candidates or specific answer candidates) to be returned as the final answer are narrowed down. For example, the basic answer candidate or unique answer candidate corresponding to the highest answer certainty of the received primary answer (identifier) is obtained from the final answer storage unit 18 (28) (step S7).
After that, the obtained final answer candidate is transmitted from the final answer means 15 to the question terminal 90 that asked the question (step S9).

If the primary answer does not satisfy the predetermined condition in step S5 (step S5; No), for example, if the answer certainty is equal to or less than a predetermined threshold, there is no suitable final answer candidate for question q1. (step S8), and transmit it from the final answer means 15 to the question terminal 90 (step S9). . Here, in step S5, if the primary response does not satisfy a predetermined condition, the final response can be associated with a special identification number and stored in the final response storage unit 18 (28). It can also be embedded as data in the program. In addition, in FIG. 5, the processing steps of steps S5 and S6 may be switched in order.

When the final answer is transmitted to the question terminal 90, various kinds of information including the history of the question and the final answer and other log data are stored in the operation history storage unit 13 (step S10). When a plurality of primary responses are presented in descending order of response certainty from the primary response system 30 (27), the operation history storage unit 13 stores all the answer certainties or up to a predetermined order. It is desirable to store the answer confidence.

As is clear from the above description, from the primary response system 30 (27), the primary response (identifier) corresponding to the question group to which the assumed question to which the input question is similar belongs, and the system functions. If so, its answer certainty is output.
By the way, the questioner thinks that he/she has asked the system user (one of the traders A to N) who has accessed the question. Therefore, depending on the content of the question, it is necessary to return different answers according to the user systems 40 of the vendors A to N that relayed the question of the question terminal 90 . For example, in the case of a question about business hours, it is necessary to answer the correct business hours for each vendor through which the question was sent. In such a case, the system user's unique answer candidate is prepared, and the identification number of the system user is acquired based on the unique attribute information of the question stored in the operation history storage unit 13, etc. By obtaining a unique answer according to the system users (traders A to N) through the question, a final answer indicating business hours for each system user is sent back.

6 to 11, hypothetical questions to be machine-learned by the primary answer system 30 (27), primary answers linked to question groups, and final answer candidates in the final answer storage unit 18 (28) Basic answers and unique answers to be stored will be explained using specific examples.
In the examples of FIGS. 6 to 11, a trader is a system user who wishes to use this question answering system, and is a trader who conducts business related to real estate transactions. In the examples of FIGS. 6 to 11, a plurality of traders (traders A to N) have contracted to use the question answering system 25 (25'), and the respective question pages of the user systems 40 of the traders A to N It is assumed that the question answering system 25 (25') can be accessed via. 6 to 9 show specific examples of question groups, and FIGS. 10 and 11 show examples of basic answer candidates and unique answer candidates stored in the final answer storage unit 18 (28). .

The question answering system 25 (25') responds to the question based on the information (communication route information: a part of the unique attribute information of the question) that indicates which user system 40 each question came from. A system user (vendor) is identified and unique answer candidates unique to the system user are acquired. If the system user (vendor) does not request specific additional information, basic answer candidates are acquired and sent to the question terminal 90 as the final answer. Although not shown in the examples of FIGS. 6 to 11, it is also possible to have a simple configuration in which only basic answers are final answers for all questions without allowing unique answers for each system user (vendor). is.

First, with reference to FIGS. 6 to 9, the assumed questions and primary answers to be learned by the primary answer system 30 (27) will be described using specific examples. Note that the following description shows a configuration example in which one question group includes a plurality of assumed questions.
FIG. 6 is a diagram exemplifying the contents of the common attributes of each question group in the question groups Q1 to QN and the relationship of the primary answers corresponding to each question group Q1 to QN on a 1:1 basis. In the following figures, the question group is appropriately indicated as "question group". In FIG. 6, the primary responses are identifiers 101 to NNN. This identifier only needs to be able to specify the primary response, and the expression format is free. Each question group Q1 to QN has a plurality of hypothetical questions grouped by the concept of common attributes for each group. Part of the question is shown as a concrete example.

FIG. 6 exemplifies that the question group Q1 is a group having the common attribute of "real estate agent" and the primary answer is the identifier "101". The question group Q2 has a common attribute of "standard for borrowing a housing loan", and the identifier of the primary answer is "102" as an example. Similarly, the question group QN has a common attribute of "points for collecting local information", and the identifier of the primary answer is "NNN".
This identifier is the same as the identifier used when creating the final answer in the question accepting and answering system 10 or 26, which will be described later, and the identifier associated with the final answer candidate, which will be described later with reference to FIGS. 10 and 11. is.

7 to 9 show specific examples of a plurality of hypothetical questions for question groups Q1, Q2, and QN. FIG. 7 specifically exemplifies a plurality of assumed questions Q1(1) to Q1(n) corresponding to "real estate agent", which is a common attribute of the question group Q1. In addition, FIG. 8 shows a specific example of a plurality of assumed questions Q2(1) to Q2(9) corresponding to the common attribute of the question group Q2, ie, "standard for borrowing a housing loan". Similarly, FIG. 9 shows specific examples of a plurality of hypothetical questions QN(1) to QN(12) corresponding to the common attribute of the question group QN, ``points for collecting local information''.

In FIG. 7, a single identifier "101" is specified for all of the n assumed questions of the question group Q1 as primary answers. That is, the primary answers to the n assumed questions in the question group Q1 having the common attribute of "real estate agent" are all the single identifier 101. FIG. Correspondingly, the final answer storage unit 18 (28) stores basic answer candidates corresponding to the identifier 101 of the primary answer, and the basic answer candidate corresponding to the identifier 101 is among the basic answer candidates. , is an answer bundle containing all answers corresponding to common attributes to multiple hypothetical questions Q1(1)-Q1(n) in the question group. Specific examples of final answer candidates stored in the final answer storage unit 18 (28) will be described later with reference to FIGS. 10 and 11. FIG.

Similarly, FIG. 8 specifically exemplifies a plurality of assumed questions Q2(1) to Q2(9) corresponding to the question group Q2, and the plurality of assumed questions Q2(1) to Q2(9) The same identifier "102" is specified as the primary response to all. Similarly, in FIG. 9, 12 assumed questions QN(1) to QN(12) related to the question group QN related to the common attribute "points when collecting information on the area" are specifically illustrated, The same identifier "NNN" is designated as the primary answer for these 12 hypothetical questions.

10 and 11 show specific examples of basic answer candidates and unique answer candidates stored in the final answer storage unit 18 (28). 10 and 11 correspond to the identifiers 101 to NNN shown in FIGS. 6 to 9, which are primary responses from the primary response system 30 (27). ing. In the final answer storage unit 18 (28), "basic answer candidates" and system user ( If one of the vendors A to N) desires, it stores a "unique answer candidate" corresponding to the desired system user.
In the columns of user (trader) identifiers in FIGS. 10 and 11, "trader A", "trader B", etc. are displayed as system user identifiers. You can freely use numbers, letters, symbols, etc. In FIGS. 10 and 11, the system user identifier is used as the information that identifies the user. Also good.

The system user (vendor) identifier is a unique identifier assigned to each of the vendors A to N, and the unique answer candidate for each vendor linked to this vendor identifier is stored in the final answer storage unit 18 (28). remembered. 10 and 11 exemplify the contents of basic answer candidates assigned as final answers. Referring to the identifiers "102" and "NNN" of the primary response in FIGS. There are no unique answer candidates corresponding to the identifiers of vendor A and vendor C in the primary response identifier "NNN" column. In this way, the setting of unique answer candidates by the system user (trader) is not essential, and can be arbitrarily set at the request of the system user (trader). Also, for all primary responses or for a specific primary response, all unique response candidates may be eliminated and only basic response candidates may be used.
The identifier of the system user (vendor) is determined from the unique attribute of the question for each question, and the identifier of the system user (vendor) in each question is determined, and the basic answer candidate corresponding to each question or the corresponding unique answer candidate is read from the final answer storage unit 18 (28).

10 and 11 exemplify specific answer contents of basic answer candidates. As can be seen from the contents of the basic answer candidates, the basic answer candidates and the unique answer candidates, which are the final answers, are relatively long answers that include all answers related to multiple assumed questions of the common attribute question group. . This is based on having the final answer be a bundle of answers containing all the answers related to multiple hypothetical questions in each question group. With this configuration, even if the number of hypothetical questions and primary answers to be machine-learned by the primary response system is reduced, it is possible to maintain the questioner's satisfaction with the final answer to the question at a certain level or higher. Become. Since the hypothetical questions are carefully selected questions that are likely to be asked, the basic answers including all answers to the hypothetical questions contain a lot of useful information for the questioner. Therefore, there is a possibility that the basic answer to one question may include the answer to the question that the questioner planned to ask next, and in this case, the questioner's satisfaction can be further improved. can.
By giving a wide range of meanings to multiple hypothetical questions in this way, it is possible to reduce the time required for creating learning data and machine learning for artificial intelligence to 1/10, 1/100 or less, and using artificial intelligence. A question answering service can be efficiently provided at a very low cost.

Also, from the primary response system 30 (27), the same primary response is always given to assumed questions belonging to the same question group, regardless of the unique attributes of the question, such as which user system the question passed through. (identifier) is output. On the other hand, in the question receiving and answering system 10 (26), in place of the basic answer candidate stored in association with the primary answer, the unique answer candidate unique to the system user is read as necessary. , the read unique answer is sent to the question terminal 90 as the final answer. Therefore, unique answer candidates are stored in association with the identifier of the primary answer and the identification information of the system user (system user identifier). A unique answer candidate is a replacement of the basic answer candidate corresponding to the identifier of the primary answer in accordance with the system user.
In this way, since the final answer candidates (basic answer candidates and unique answer candidates) stored in the final answer storage unit 18 (28) are stored data separated from artificial intelligence, machine learning for artificial intelligence is possible. can be changed or added as appropriate regardless of Therefore, maintenance work such as changing or adding final answer candidates stored in the final answer storage unit 18 (28) does not require advanced expert knowledge of artificial intelligence. However, the basic answer candidate corresponds to the first answer of the first answer system, and is the basis of the final answer and is the answer content common to each system user, so that the system user cannot change it. preferably limited.

As is clear from the above description, the question answering system 25 (25') according to the present invention has accumulated experience and achievements in order to efficiently answer questions from end-users or potential end-users. Data content such as hypothetical questions and final answer candidates, as well as equipment for automatically providing appropriate answers in response to questions, such as advanced computer systems such as artificial intelligence systems and databases, etc. It is possible for multiple system users (traders, etc.) to jointly use the system. By enabling the construction of a question answering system that enables such shared use, it is possible not only to greatly reduce the cost of constructing a question answering system when introducing it for the first time, but also to shorten the system introduction period.
In addition, if a vendor who shares the same data and system introduces it later, the construction of the data and system environment has already been completed, so you can safely introduce an excellent Q&A system in an extremely short period of time. has the effect of In addition, since many users jointly use the question answering system, it is possible to receive a wide variety of questions in a short period of time, which cannot be obtained by a single company. Therefore, it becomes possible to grow into a better system in a short period of time.

The question receiving and answering system 10 (26) according to the present invention is provided with the maintenance means 16 as described above, and the initial setting means 21 allows the primary answer system 30 (27) and the final answer storage section 18 (28) Make initial settings. The initial setting procedure of the primary response system 30 (27) and the final response storage unit 18 (28) will be described with reference to FIG.

FIG. 12 is a flow chart illustrating the procedure of basic work for initial setting of the primary response system 30 (27) and the final response storage unit 18 (28). In order to build the system, first, the assumed question, the common attribute of the assumed question, the question group, the primary answer, the final answer (basic answer, unique answer), and the unique attribute of the question are determined (step S21).
For this purpose, the work of the system user (industry) is analyzed, a plurality of assumed questions are prepared, and answers to all the assumed questions are prepared. As for the assumed questions, a plurality of assumed questions having a common attribute are grouped into one question group, and the primary answer is an identifier corresponding to the question group on a 1:1 basis. Of the final answer candidates, the basic answer candidates have contents that cover all answers related to a plurality of assumed questions in one question group. Also, the unique attribute information is determined, and the range and contents of the unique answer candidates linked with the unique attribute information are determined. At that time, language information can also be included in the unique attribute information. In that case, the final answer candidates that differ from the expected questions in various languages are prepared as necessary. Such final answer candidates are created for all question groups for each question group.

Next, let the primary answer system 30 (27) perform machine learning on all the assumed questions included in each question group Q1 to QN and the primary answers (identifiers) corresponding to each question group and 1:1. (Step S22). Further, the final answer storage unit 18 (28) associates the identifier of the primary answer with the basic answer candidate and stores them (step S23). As described above, in step S23, not only the basic answer candidate for one identifier corresponding to the primary answer, but also unique answer candidates different for each system user based on the common basic answer candidate are applied to each first answer. It may be configured to be stored in association with the primary response and the identifier of the system user. The order of step S22 and step S23 may be reversed or may be performed at the same time.

The question answering system 25 (25') according to the present invention comprises maintenance means 16 as described above. With this maintenance means 16, basic answer candidates and It is possible to modify or add unique answer candidates.

The question answering system 25 (25') of the present invention maintains the machine learning of the primary answer system 30 (27) and the final answer candidates stored in the final answer storage unit 18 (28) from the results of monitoring. This allows us to grow continuously. In addition, regular checking of questions with low answer confidence leads to opportunities for each user to learn about the "needs of existing and potential customers" that they themselves are not aware of, and the question answering system simply answers questions. In addition to the original purpose of the system 25 (25'), it can be a means of collecting customer and market information necessary for business growth. In addition, as described below, basic answer candidates and unique answer candidates can be maintained under certain conditions even by system users who do not have knowledge of artificial intelligence. System users themselves can continuously improve and grow the question answering system.
In addition, since this question answering system can be used jointly by a plurality of system users, it is possible to receive a much greater variety of questions than when a single company uses it. Therefore, during maintenance of the entire question-answering system 25 (25') by the system provider, the expected questions and basic answer candidates are modified and improved to deal with a wide variety of problems, and the growth speed of the question-answering system increases. It is expected. As for the growth of this question answering system, all co-users can benefit from it.

13 and 14, the procedure for maintenance of the primary response system 30 (27) and the final response storage unit 18 (28) by the system user (trader) or system administrator will be described. First, a procedure for correcting/changing the final answer storage unit 18 (28) by the system user (trader) will be described.

FIG. 13 is a flow chart showing a processing procedure for a system user (trader) to correct/change the final reply unique to the company.
The system user (vendor) uses the management terminal of the system user connected to the question answering system 25 (25') via any communication network to maintain the question answering system 25 (25'). 16 (see Figures 3 and 4), and from records such as access logs where questions were asked to the system user, distribution of answer confidence of questions and primary answers, average answer confidence, etc. It is possible to check the status of other final answers, and to consider corrections, changes, additions, etc. to the content of final answer candidates (basic answers and unique answers). If there is a case where an appropriate final answer could not be given to the question from the question terminal, or a case where the final answer was given but the answer confidence of the first answer was low, the system user (contractor) The user can modify or add the final answer candidate stored in the final answer storage part within the scope of authority given to him/herself or by requesting the system administrator.

As shown in FIG. 13, the system user (trader) regularly or irregularly checks and analyzes the access log (step S31). In that case, if you wish to modify or change the final answer (basic answer and unique answer) stored in the final answer storage unit 18 (28) (step S32; Yes), then correct/change the final answer. It is checked whether or not there is an authority to equalize (step S33). Although not shown in FIG. 13, when the system user (dealer) browses the operation history information and analyzes the log, etc. (step S31), it is necessary to check whether the person in charge has the access authority in advance. It is desirable to check whether If the system user (dealer) does not have the authority to correct or change, the maintenance process is terminated without permitting the correction or change (step S33: No). If the authority to modify, change, etc. is granted (step S33: Yes), the final answer candidate is modified ( (including changes or additions) (step S34).

For example, when the question terminal 90 asks the question "I want to know about business hours" as the question q1, and there is already an answer specific to the system user, information specific to the company, such as temporary closure, can be sent to the existing system. It can be added to the unique answer candidates. If the system user (vendor) wishes to make corrections that exceed the authority of the system user based on the results of analysis of log data, etc. ), ask the system administrator of the question-answering system to consider adding assumed questions to existing question groups, adding new question groups with common attributes, or modifying or adding basic answer candidates. do.

FIG. 14 is a flowchart illustrating a maintenance processing procedure for the entire question answering system 25 (25') by the system administrator. This maintenance processing procedure is a processing procedure performed via the system management terminal 50 by a system administrator who manages the operation of the entire question answering system.
When starting maintenance (step S41), first, it is confirmed whether or not there is access authority to this operation history (step S42), and if there is no access authority, this processing flow ends (step S42; No). If the access authority is granted (step S42; Yes), the access status of the entire system stored in the operation history storage unit 13, log data of questions and answers, and analytical data such as statistical data are checked and analyzed. (Step S43).

If the final answer candidates in the final answer storage unit 18 (28) are to be corrected or changed (including additions, the same shall apply hereinafter) (step S44; Yes), it is confirmed whether or not the authority for such correction or change is granted. (step S45). If there is an authority to modify or change (step S45: Yes), the basic answer/specific answer is changed or a new supplier's specific answer is added (step S46). If the user does not have the authority to modify, change, etc., the process is terminated (step S45: No). Of the final answer candidates, only the system administrator is permitted to modify or change the basic answer candidates, and the system administrator is not permitted to modify or change the unique answer candidates. It may also be configured so that only the user has the authority to modify, change, etc.
This figure shows an example in which the system administrator has separate authority for final answer candidates and machine learning, and both authorities may be covered by a single authority.

If the final answer candidate is not to be corrected or changed (step S44; No), or if the final answer candidate in the final answer storage unit 18 (28) has been corrected or changed (step S46), then the artificial intelligence system It is confirmed whether or not to perform machine learning by adding hypothetical questions to a certain primary answer system 30 (27) (step S47). If there is no need for additional machine learning (step S47; No), this maintenance process is terminated. If additional machine learning is to be performed (step S47; Yes), it is confirmed whether the primary response system 30 (27) has the authority to modify or change the assumed question or primary response (identifier). (step S48). If you have the authority to modify or change (step S48; Yes), make the desired additions, modifications, or deletions to the assumed questions and primary answers, and let the primary answer system perform machine learning (step S49). end the maintenance process. If there is no correction authority (step S47; No), the maintenance process is terminated without performing machine learning.

The maintenance processing (FIG. 13) by the system user (trader) and the maintenance processing (FIG. 14) by the system administrator described above are, for example, based on the following rules to update and change data. As a result, the question answering system 25 (25') can be evolved step by step into a system with a high degree of certainty of answering, and can be reliably grown.

<Example of handling rules when answer confidence is low>
(1) If the content of the question is completely unknown and there is no corresponding final answer, the identifier pair of the question and answer is added to the primary answer system 30 (27), and the final answer is The final answer is also added to the answer storage unit 18 (28). This maintenance is performed by the system administrator.
(2) If the content of the question is unknown, but there is an appropriate existing final answer, the primary answer system 30 (27) stores the identifier pair of the question and the existing final answer. Add In this case as well, the system administrator performs this.

<An example of a response rule when the response confidence exceeds the reference value, but the content of the response is not appropriate>
(1) If the content of the question is already known but the content of the final answer is not valid, the system administrator corrects only the final answer data in the final answer storage unit 18 (28).
(2) If it is necessary to correct the basic answer, which is a common answer, for all related traders, the system administrator can correct the basic answer in the final answer storage unit 18 (28). .
(3) If the basic response, which is a common response, does not match the content of some system users (vendors), the system administrator of the system users (vendors) will provide the final response corresponding to their company's vendor identification number. You can copy the basic answer to the area and modify the contents of the basic answer. At this time, when correcting the final answer of only one company that does not affect the common answer, addition/correction can be made for each vendor identifier.

<Correspondence rules when it is assumed that answer accuracy will be improved by dividing the question group>
(1) If there is an answer identifier with a large number of questions and answers, and it is assumed that dividing the question group and the final answer by narrowing the range of answers will improve user satisfaction, Divide the currently registered question group and final answer into one. This maintenance processing is performed by the system administrator.
(2) Even if the length of the answer to the question becomes too long, it is desirable to divide the question group and the final answer in the same way as when the number of questions is large. This maintenance processing is performed by the system administrator.

<Support for multiple languages>
In the above-described embodiments and specific examples, the question answering system of the present invention has been explained by exemplifying the assumed question and the final answer using Japanese as the question language and answer language.
The question answering system 25 (25') according to the present invention can respond to questions in many kinds of languages by grouping assumed questions and using unique attribute information of the questions. An example of the configuration of the question answering system 25 (25') according to the present invention for responding to questions and answers in a plurality of languages will be described below with specific examples of assumed questions and final answer candidates.

When the native language of the questioner and the respondent is different, smooth communication is difficult unless either the questioner or the respondent is familiar with the other's native language. In such a situation, if there is a question answering system that can accept questions in the native language of the questioner, the questioner will be able to create a clear question sentence using correct documents. Being able to ask questions in the native language in this way improves the quality of questions, so the probability of providing appropriate answers from the question-answering system is greatly improved.

Therefore, in the question answering system, it is desirable to have a mechanism for accepting questions in various languages. In general, the ability to read a document is relatively high compared to the ability to create a document, even if it is not the native language. For example, from the viewpoint of cost-effectiveness, while allowing a wide variety of question languages, the answer language should be narrowed down to the major languages of the country where the question-answering system is installed, as well as a few languages that are widely used around the world, such as English and Chinese. is also possible.

Such a multilingual question answering system can greatly reduce the human burden in various consultation and guidance services in administrative agencies, universities, hospitals, and the like. For example, in local governments, answering questions in various languages about simple questions such as where to find the counters and facilities for various procedures such as the counter for procedures for moving in, the counter for providing resident cards, and the counter for tax payment in local governments. Even if it is possible to automatically respond via the system, it is possible to greatly reduce the individual response effort by humans.
Especially in the first consultation or in the early stages of consultation, many of the questions are formal or simple and basic questions. It will be a very useful system for both users and institutions.
In addition, the fact that the questioner can use their smartphone from a remote location to ask questions in advance to such a question-answering system will make it easier for the questioner and the user institution to ask questions compared to the case where it was necessary to ask questions on site. It will be a very useful system for both parties.

In the question answering system 25 (25') according to the present invention, a question answering system 25 (25') capable of responding to many kinds of languages can be constructed by simply adding a simple configuration to the basic configuration of the present invention. be able to.
In the question answering system 25 (25') according to the present invention, assumed questions with the same common attribute are configured to belong to the same question group even if the language of the question differs. Supports questions and answers in multiple languages by recognizing them as part of the question's unique attributes.
For this reason, a plurality of hypothetical questions that differ in language but have the same meaning and content are assumed to belong to the same question group with the same common attribute, and are machine-learned by the primary answer system 30 (27). At the same time, final answer candidates (basic answer candidates or specific answer candidates) for each type of answer language are stored in the final answer storage unit 18 (28). Answers in the default language set in the question answering system 25 (25') are registered as basic answer candidates, and answers in other languages are registered as unique answer candidates.

When responding to questions in multiple languages, it is necessary to extract a correct primary answer according to the language of the question and to answer in an appropriate language corresponding to the language of the question. To establish the question language and answer language, the questioner can be prompted to enter the desired question language and answer language. As the simplest method, when the question answering system 25 (25') is accessed from the question terminal 90, a language selection button for specifying the type of question language and answer language is provided from the question input screen of the question terminal. can be done. In this case, it is possible to select a desired language type using a pull-down method or a radio button method, so that only the types of languages that can be handled by the question answering system can be specified. desirable.
Also, if the question receiving and answering system 10 (26) or the primary answering system 30 (27) has a function to infer the type of question language from the form of the question text (character code bit configuration, etc.) It is also possible to determine the type of language from the question text by the question receiving and answering system 10 (26) or the primary answering system 30 (27). Furthermore, the type of question language can be obtained from common information required for communication between the question terminal and the question answering system. Therefore, it is possible to assume the question language without the questioner's input. Regarding the answer language, it is possible to deal with it if it is initially set so that it is determined according to a predetermined rule according to the question language.

The desired question language and type of answer language are sent to the question receiving and answering system 10 (26) together with the question, and are stored in, for example, the operation history storage unit 13 as part of the unique attributes of the question. Here, configuring so that the reply language is automatically set to the same language as the input language will lead to improved convenience for the user. However, in view of the fact that document reading comprehension ability is much higher than document creation ability, from the viewpoint of reducing the load at the time of system introduction and from the viewpoint of reducing the response load of the question answering system 25 (25') , the type of answer language may be set to only English or Japanese.

As described above, first, it is necessary to receive questions in multiple languages from the primary response system 30 (27) and obtain appropriate primary responses regardless of the language of the questions. For that purpose, multiple hypothetical questions with the same common attribute are created according to the type of language, and hypothetical questions in multiple languages with one common attribute are treated as one question group, and correspond 1:1 with each question group. Machine learning is applied to the primary response system so that the primary response to be obtained can be obtained.
For example, question groups Q1 to QN are created in which assumed questions in a plurality of languages, such as Japanese questions, English questions, and Chinese questions, having common attributes are assumed to belong to the same question group. Here, the hypothetical questions for each language having the same common attributes may be hypothetical questions with the same meaning and content, which are direct translations or paraphrases of hypothetical questions created in one language into other languages. It may be a hypothetical question with almost the same meaning in consideration of language culture. Next, the question groups Q1 to QN containing assumed questions in various languages are machine-learned so as to obtain primary answers corresponding to the question groups Q1 to QN on a 1:1 basis. As a result, from the primary response system 30 (27), the same single primary response (identifier) is obtained for the assumed questions belonging to the same question group, regardless of the type of question language.

In the question receiving and answering system 10 (26), the final answer candidates in various languages are stored in the final answer storage section 18 in correspondence with the primary answer in a one-to-one correspondence with each answer language. Therefore, when the primary response is received from the primary response system 30 (27), based on the identifier of the primary response and the language information (identifier) of the question, the final response candidate in the appropriate response language can be read out and the final answer can be returned to the question terminal 90 .

It is also possible to combine with unique answer candidates other than language, such as the vendor identifier described above. By preparing basic answer candidates corresponding to the first response and unique answer candidates that differ for each system user (contractor) according to the type of language, the system can be used in the appropriate language according to the question language. It is possible to read the unique answer candidate for each person and use it as the final answer. Whether or not to prepare a unique answer candidate corresponding to a combination of multiple types of unique attributes as the final answer can be appropriately set as required. Unique answer candidates are not necessarily required when one question-answering system is not used by a plurality of institutions, such as in cases where the system is used alone in an administrative institution, university, or the like.

When the question receiving and answering system 10 (26) receives a question from the question terminal 90, the receiving means 12 acquires the language information of the question from the unique attribute information of the question, and stores it in the operation history storage unit 13 or the like. The storage location is not limited to the operation history storage unit 13, and a separate location for storing individual attribute information of questions may be provided. When the question is received, the primary answer inquiry means 14 requests the primary answer system 30 (27) for the primary answer corresponding to the question. In this case, when the primary response system 30 (27) requests the language information of the question, the question text and its language information are sent from the question reception and answer system 10 (26), and the language information of the question is not requested. In that case, only the questionnaire will be sent. When the question receiving and answering system 10 (26) receives the primary answer from the primary answering system 30 (27), it acquires the identifier of the answer language based on the language information stored in the operation history storage unit 13. . After that, based on the identifier of the primary answer and the answer language information (identifier), the final answer candidate in the appropriate answer language is obtained.

With reference to FIGS. 15 to 19, specific examples of question groups, assumed questions, and final answers when constructing a question answering system capable of answering questions in various languages are assumed to be questions and answers in administrative agencies. will be used to explain.
15 to 19 show question groups Q1 to QN, assumed questions Q1(1) to QN(8) in Japanese, English, and Chinese in each question group, and final answer candidates in Japanese, English, and Chinese. Specific examples of (basic answer candidates and unique answer candidates) are shown in simplified form.

Assuming that the question answering system according to the present invention is introduced at a city hall, FIG. shows a specific example of the first response corresponding to . In the example of FIG. 15, the common attribute of question group Q1 is "transfer procedure", the common attribute of question group Q2 is "acquisition of resident card", and the common attribute of question group QN is "payment of tax". are "101" to "NNN", respectively. FIGS. 16 to 18 show assumed questions and primary answers (identifiers) included in question groups Q1, Q2, and QN, respectively, and FIG. 19 shows specific examples of final answer candidates.
To simplify the explanation, FIG. 19 shows only the final answer for each language, but it is also possible to provide unique answers other than languages for each type of language.

FIG. 16 shows an example in which the number of assumed questions for Japanese is four, English is three, and Chinese is two. The number of assumed questions for each language may be the same, but the number of assumed questions for each language does not necessarily have to be the same as shown in FIG. In the initial stage of introducing a Q&A system, it is better to prioritize being able to answer basic questions and be able to respond to questions in English or Chinese rather than preparing a large number of assumed questions. , it often leads to improvement of convenience not only for the questioner but also for the user institution. This is because foreigners from different cultures often do not understand basic matters, so it is expected that there will be a need to respond to foreigners who ask questions. The question may have the same meaning in each language, or may have different meanings and nuances for each language in accordance with customs, culture, and the like.

In question group Q2 shown in FIG. 17, "acquisition of resident's card" is selected as the common attribute, and the primary answer of question group Q2 is "102." As hypothetical questions belonging to the question group Q2, FIG. 17 shows specific examples of hypothetical questions in multiple languages. In question group QN, "payment of tax" is selected as the common attribute, and the primary answer for question group QN is "NNN". Further, as the assumed questions of the question group QN, FIG. 18 shows specific examples of assumed questions in a plurality of languages.

FIGS. 16, 17 and 18 exemplify concrete question sentences in Japanese and English, and Chinese is schematically shown using Japanese.
As can be seen from the figure, for example, the primary answer to the questions in the question group Q1, as shown in FIG. . In other words, although the question sentences are assumed questions in Japanese, English, and Chinese, the primary answer to the question group Q1 is only the identifier "101" regardless of the assumed questions in any language. This is the same for question groups Q2 and QN as shown in FIGS. The answer for Group QN is "NNN".
In the primary answer system, machine learning is performed on all assumed questions and primary answers belonging to each of these question groups, and the primary answers from the primary answer system are, regardless of language, All of them are configured to have the same answers "101" to "NNN".

FIG. 19 shows specific examples of primary answers, language identifiers that are part of the unique attributes of questions, and final answer candidates stored in the final answer storage unit 18 (28). In the example of FIG. 19, only final answer candidates in various languages are shown, without showing unique answer candidates other than language information, in order to facilitate understanding of handling of various languages. However, as in the examples described with reference to FIGS. 5 to 11, in addition to the final answer candidates for each language, unique answer candidates that differ for each system user (trader) are prepared for each type of answer language. By doing so, it is possible to configure so that the final answer in an appropriate answer language is sent back to the question terminal 90 .

When the question answering system acquires the primary answer, the final answer means 15 acquires the language identifier of the answer language from the unique attribute information of the question stored in the operation history storage unit 13, A corresponding final answer candidate is read from the final answer storage unit 18 . For example, if the identifier "101" is acquired as the primary answer and the language identifier of the answer language is English, the final answer candidate in English as shown in FIG. and sent to the question terminal 90 as the final answer.
In the example shown in FIG. 19, since there are no unique answer candidates for each system user (vendor), the English final answer candidate is returned as it is as the final answer. By providing correspondingly, it is possible to return the English unique answer candidate as the final answer.

In the above description, an example was shown in which the questioner individually specifies the question language and the answer language. However, as described above, the information on the question language can also be automatically determined by the question reception/answer system 10 (26) or the primary answer system 30 (27) from the form of the question sentence. In addition, even without input from the questioner, it is possible to obtain common information necessary for communication between the question terminal and the question answering system based on the OS of the question terminal, language information set in the web browser, etc. be. Therefore, the question language and the answer language can be set by the user for each question. It is also possible to answer in the language specified by the questioner, or to have a configuration in which specification by the questioner and initial setting are combined.

Recognition of language information includes specification of the language of the questioner, language recognition by the automatic discrimination function from the question sentence by the question reception and answer system 10 (26) or the primary answer system 30 (27), or Web browser, OS, etc. Regarding language recognition based on the set language information, it is preferable to determine the question language with priority. The answer language can also be set to be the same language as the question language, or can be initially set to answer in a language different from the question language. The answer language determination rule can be appropriately adopted based on the questioner's convenience, system efficiency, and the like.

FIG. 20 shows an example of a processing procedure when different types of question languages are used.
Steps S1 to S4 and S9 to S10 are the same as the steps shown in FIG. 5, although the information handled by the unique attribute information of the question is slightly different. Simplify the description. As explained in FIG. 5, when the question q1 is received from the question terminal 90 (step S1), the unique attribute information of the question is acquired (language information of the question, system user information, etc.) and stored in, for example, operation history storage. It is stored in the unit 13 (step S2). This step S2 is different from step S2 in FIG. 5 in that the language information of the question and the like are acquired and stored as unique attribute information. After that, the question received from the question terminal 90 is transmitted to the primary response system 30 (27) to request the primary response (step S3). Upon receiving the question q1, the primary response system 30 (27) extracts and outputs one or more primary responses corresponding to the question. 27) receives the primary response (step S4). The processing steps up to this point are the same as those described in FIG. 5, except that the language information of the question is acquired and stored as the unique attribute information of the question in step S2.

As described above, even if the language information of the question is acquired by input from the questioner, the question reception and answer system 10 ( 26) may be automatically obtained. When the answer language is specified by the questioner, it is preferable that the answer language information is also stored as one of the language information in step S2. These language information are stored as predetermined language identifiers.
In the following explanation, the language of the question is Chinese, the questioner did not specify anything as the answer language, and if the question language is Chinese, the final answer is also the same language as the question. It will be explained as if the answer is in Japanese.
In addition, in order to reduce the number of languages for answers to questions in multiple languages, or if the final answer in Chinese is not prepared in the final answer storage part because the preparation of the answer in Chinese is not in time, the answer will be in English. It is also possible to configure In this case, the final answer storage unit 18 (28) stores the final answer candidate in English in association with the identification information of Chinese, and when the answer language is Chinese, the final answer Various measures are possible, such as a configuration in which initial settings are made so that the English final answer candidates in the storage unit 18 (28) are read out.

When the primary response is received from the primary response system 30 (27) (step S4), the identifier of the response language and the identifier of the system user stored in the operation history storage unit 13 and the like are acquired (step S55). ). At present, the question language is Chinese, and the questioner has not specified the answer language, so according to the initial setting, the Chinese identifier is obtained as the answer language. At the same time, it is checked whether the primary response satisfies a predetermined condition (step S56).

If the primary response satisfies a predetermined condition such as having a certain degree of certainty of response (S56: Yes), the final response storage unit 18 will respond based on the identifier of the primary response and the identifier of the response language. A final answer candidate corresponding to the language is read (step S57). Here, when unique answer candidates other than language are stored in the final answer storage unit 18 (28), the unique answer candidates corresponding to the system user etc. are read based on the identifier of the system user. .
For example, referring to the examples of FIGS. 15 to 19, if there is a Chinese question “I came to pick up my resident card”, the assumed question Q2(7) shown in FIG. , the question receiving and answering system 10 (26) receives the primary answer identifier "102" (step S4) from the primary answering system 30. Also, since the answer language information is not specified at the time of questioning, the language identifier of Chinese is acquired as the answer language in step S55. After that, as shown in FIG. 19, based on the primary response identifier "102" and the Chinese language identifier, the Chinese language of the primary response identifier "102" is retrieved from the final response storage unit 18 (28). A final answer candidate is read (step S57). Although FIG. 19 does not show specific documents of the final answer candidate in Chinese, the final answer in Chinese is actually stored.

In the example shown in FIG. 19, since no unique answer candidates other than the corresponding language are prepared, the final answer shown in FIG. 19 becomes the final answer (step S57) and is transmitted to the question terminal 90 (step S9 ).
Here, if there is no unique answer candidate other than the corresponding Chinese language, but there is a corresponding unique answer candidate in a predetermined language (for example, Japanese), the unique answer candidate in that language (Japanese) is the final answer. It may be configured to reply as Finally, the final reply sent to the questioner and other information are stored in the operation history storage unit 13 (step S10).

If the primary answer does not satisfy the predetermined conditions (S56: No), the final answer candidate prepared for each answer language, such as "Please ask a little more specific question." is obtained from the final answer storage unit 18 (28) based on the answer language information (step S58) and transmitted to the question terminal 90 (step S9). Alternatively, a similar final answer may be generated by a program (step S58) and transmitted to the question terminal 90 (step S9).
Here, instead of preparing individual final answer candidates for each answer language, for example, multiple documents stating "Please ask a more specific question" in all answer languages are prepared as one final answer candidate. However, if the primary answer does not satisfy a predetermined condition, the final answer candidates can be read out and sent back to all the questions.
After that, various information from the question to the final answer regarding the question is stored in the operation history storage unit (step S10).

<Question Answering System According to Another Embodiment of the Present Invention: Entity Function>
Furthermore, another embodiment of the present invention will be described.
If the primary answer system 30 (27) equipped with artificial intelligence has a function of notifying an object (entity) included in a question, a predetermined entity is registered in advance in the primary answer system. By doing so, it is possible to greatly reduce the number of assumed questions for machine learning.
If the primary response system 30 (27) has a function of outputting the entity included in the question, by registering the entity in advance, the primary response system can , can receive the entities contained in the question.
In order to utilize the output function of this entity, create a question group with "question intention" as a common attribute in the assumed question and multiple assumed questions with the target (entity) included in the intention of the question as a variable, The necessary entities are also registered in the primary response system 30 (27). In this way, by registering multiple entities corresponding to each question group and preparing multiple assumed questions with entities as variables, the number of registered entities multiplied by the number of assumed questions created Only assumed questions are prepared, and the number of assumed questions actually machine-learned can be greatly reduced. Here, entity registration can be performed as a system-common entity instead of registering the entity for each question group.

With reference to FIGS. 21 to 27, a method of envisioned questions and final answers in a question answering system using entities will be described with reference to specific examples. 21 to 27 are examples of a question answering system 25 (25') in a city hall. FIG. 21 exemplifies common attributes of assumed questions belonging to each question group Q1 to QN and identifiers of primary answers. In the example shown in FIG. 21, question groups Q1 to QN-2 are composed of hypothetical questions having the same common attributes as before.

In each question group Q1 to QN-2, the assumed questions are also assumed questions similar to those described so far, and the primary answers corresponding to each question group, as partially illustrated in FIGS. 22 to 26. 101” to “NNN-2” are prepared. Therefore, when the primary answer system 30 (27) receives a question and the question corresponds to any of the assumed questions in the question groups Q1 to QN-2, the corresponding primary answers "101" to "NNN -2” is provided from the primary response system 30 (27) to the question reception and response system 10 (26). When the question receiving and answering system 10 (26) receives one of "101" to "NNN-2" as the primary response, the final response corresponding to the identifier of the received primary response as illustrated in FIG. Acquire answer candidates (basic answer candidates or unique answer candidates) and send them back to the question answering system.

On the other hand, the common attributes of question groups QN-1 and QN in FIG. entity) is not specified. In addition, as shown in FIGS. 25 and 26, the assumed questions included in each question group QN-1 and QN include specific questions such as which window you want to go to or what kind of documents you want to obtain. The target is not included, and the variable is "xxx". That is, in the question groups QN-1 and QN, the common attribute is "intention to do something", and the object (entity) is the "object of what you want to do". ", or desired "document") is a variable. Such question groups QN-1 and QN are hereinafter referred to as "intent question groups".
On the other hand, if the question from the questioner is to request a destination or to obtain documents, the question must include a specific object. Therefore, when the primary response system outputs "NNN-1" or "NNN" as the primary response, the entity included in the question should also be output at the same time.

In the question reception and answer system, when the intent (intention) question group QN-1 or the identifier "NNN-1" or "NNN" corresponding to QN is received, based on the intent (intention) included in the question must create and submit an appropriate final candidate answer that corresponds to the question.
For this reason, the question receiving and answering system has a correspondence table indicating final answer identifiers for selecting final answers corresponding to registered entities, as shown in FIG. to obtain the identifier of the final answer candidate that stores the final answer candidate corresponding to the entity. The identifier of this final answer candidate designates these identifiers when corresponding final answer candidates exist in the identifiers "101" to "NNN-2" corresponding to the ordinary question groups Q1 to QN-2. If there is no suitable final answer for the identifiers "101" to "NNN-2", a separate final answer identifier and final answer candidates may be provided.

If the primary answer is the primary answer "NNN-1" or "NNN" corresponding to the intent question group, the question receiving and answering system receives the first answer received from the primary answering system. Obtain the identifier of the corresponding final answer candidate from the next answer (identifier) and the identifier of the entity, and obtain the final answer.
For example, it is assumed that the question from the question terminal is "I came here for moving-in procedures." Since there are no hypothetical questions directly applicable to this question in question groups QN-1 and QN, we assume that the intent question group QN-1(1) is the primary answer "NNN-1". In that case, the identifier of the primary answer with the highest answer certainty is "NNN-1" and the entity "transfer procedure" is provided. When the question receiving and answering system receives the primary answer identifier "NNN-1" and the entity identifier "transfer procedure", it acquires the final answer candidate identifier "101" from the correspondence table illustrated in FIG. The final answer candidate corresponding to the identifier "101" of the primary answer shown in is read out and transmitted to the question terminal as the final candidate.

Although the primary answer "NNN-1" or "NNN" corresponding to the intent question group is received, if the entity does not exist, the final answer candidate cannot be obtained. As shown, an answer such as "I didn't know which window I wanted to go to" or "I didn't know what kind of document was required" is sent back to the question terminal as a final answer. In such a case, it is assumed that a window or necessary documents that are not registered as an entity are designated as questions from the question terminal. By collecting and analyzing such log information and changing the assumed questions and registration of entities, it is possible to increase the rate of correct answers to questions and to evolve the question answering system 25 (25').

In the above explanation, as the final answer candidates, there are examples in which "basic answer candidates" and "specific answer candidates" are used as the final answers, "basic answer candidates and unique answer candidates in various languages" are used as final answers, and "second Although an example in which "a basic answer candidate obtained from a combination of a primary answer (identifier) and an entity identifier" is the final answer, other combinations are also possible, and all configurations obtained by combining the described configurations with each other also belong to the scope of the present invention.

In the embodiment shown in the above description, only examples of question groups by natural language processing are shown, but the questions handled by the question answering system 25 (25') according to the present invention are questions based on natural language format. Not limited. The question answering system 25 (25') can also be configured to incorporate an image analysis and image recognition system, receive image data from the question terminal, and transmit a final answer related to the image. A response to a question using an image is, for example, when image data of a tangible object such as a specific building, structure, animal, or person is received as a question from a question terminal, information about the image data is returned as a final answer. It is a thing.
Specifically, it is possible to adopt a configuration in which, upon receiving as a question a photograph of the appearance of equipment such as a kitchen handled by a remodeling company, information regarding this equipment is provided as a primary answer. The question answering system 25 (25') can return an appropriate answer for each trader to the questioner by registering the information about the equipment as a unique answer for the company.

Further, by receiving the image data from the question terminal, it is possible to configure so that the answer of the information about the image is returned in a language suitable for the questioner.
As a specific application, for example, it can be used as a tourist guide. In this case, it is possible to receive an image of the destination from the inquiry terminal and provide guidance on how to get to the destination (such as how to change trains) by searching for image recognition technology and means of transportation. By acquiring language information from photos and communication information, it is possible to configure so that the description of the transit method is returned in an appropriate language.
Similarly, explanations of works in art museums and the like can be provided in various languages.

10, 26 Question Receiving and Answering System 11 Communication Interface 12 Question Receiving Means 13 Operation History Storage Part 14 Primary Answer Inquiry Means 15 Final Answer Means 16 Maintenance Means 18, 28 Final Answer Storage Part (Final Answer Database)
25 25' question answering system 27, 30 primary answer system 35 final answer control unit 36 unique attribute information acquisition unit 37 final answer acquisition unit 38 final answer transmission unit 40 user system 90 question terminal 91 communication network (complex including LAN Network: wired or unrelated)

Claims (13)

A question receiving and answering system using a computer that receives image data sent from a question terminal as a question based on image data and returns a final answer to the question based on the image data to the question terminal; a question answering system separately provided with a primary answer system for replying a primary answer to a question based on the image data,
The primary response system is
Receiving a reply request for the primary answer to the question by the image data from the question reception and answer system, inputting the question by the image data received from the question reception and answer system to the artificial intelligence, and receiving the question from the artificial intelligence interface means for transmitting a primary answer to a question based on the image data of to the question acceptance and answer system;
The image data from the question receiving and answering system in response to a request from the question receiving and answering system by machine learning a plurality of assumed questions and a predetermined identifier as a primary answer corresponding to the assumed question. and an artificial intelligence that outputs the primary answer to the question by as the identifier,
The question reception and answer system is
a question accepting means for accepting questions based on the image data from the question terminal;
A first order for transmitting a question based on the image data accepted by the question accepting means to the primary answer system, requesting the primary answer, and receiving the primary answer from the primary answer system answer inquiry means;
a final answer storage unit that associates 1:1 with the identifier machine-learned by the artificial intelligence and stores a final answer candidate corresponding to the identifier;
based on the primary response received from the primary response system, the final response candidate corresponding to the identifier of the primary response is read from the final response storage unit, and the read final response candidate is stored a final answer means for returning a final answer to the question terminal based on the question answering system. A question receiving and answering system using a computer that accepts questions based on image data from a question terminal, and artificial intelligence, and machine-learns a plurality of assumed questions and a predetermined identifier as a primary answer corresponding to the assumed questions. The question receiving and answering system in the question answering system provided separately from the primary answering system that outputs the primary answer to the question by the image data from the question receiving and answering system as the identifier,
a question accepting means for accepting questions based on the image data from the question terminal;
A first order for transmitting a question based on the image data accepted by the question accepting means to the primary answer system, requesting the primary answer, and receiving the primary answer from the primary answer system answer inquiry means;
a final answer storage unit that associates 1:1 with the identifier machine-learned in the primary answer system and stores a final answer candidate corresponding to the identifier;
Based on the primary response received from the primary response system, the final response candidate corresponding to the identifier of the primary response is read out from the final response storage unit, and the read out final response candidate is stored. a final answer means for returning a final answer to the question terminal based on
Question acceptance and answer system. The artificial intelligence of the first-order answer system is 1:1 with a large number of the assumed questions that constitute a plurality of question groups, with one or more of the assumed questions having a common attribute as one group, and the question groups. The question answering system according to claim 1 or the question answering system according to claim 2, characterized in that machine learning is performed so that one identifier corresponding to is the primary answer of the assumed question included in each question group. question reception answer system. The question receiving means further acquires unique attribute information of the question by the image data accompanying the question by the image data from the question terminal,
4. The question answering system according to claim 1 or 3, or claim 2 or 3, wherein the final answer storage unit stores final answer candidates corresponding to the unique attribute information as needed. Question acceptance and answer system described in . The question receiving means acquires identification information of a user system through which the question based on the image data is passed, based on the acquired unique attribute information,
The final answer storage unit stores, as each final answer candidate, a basic answer that is the base of the final answer stored in association with each identifier, or, if necessary, each identifier and identification information of the user system. Each of the basic answers stored in association with each other and unique answers having different contents are stored,
The final answer means reads the basic answer or the unique answer from the final answer storage unit based on the identifier of the primary answer and the identification information of the user system corresponding to the question based on the image data. 5. The question answering system according to claim 4 or the question receiving and answering system according to claim 4, wherein the read basic answer or the unique answer is transmitted to the question terminal as a final answer. Said question receiving means acquires identification information of said user system as said unique attribute information from communication route information through which said question receiving and answering system is reached or information given in a communication process. The question answering system according to claim 5 or the question receiving and answering system according to claim 5. The primary answer system extracts a primary answer to the question based on the image data and outputs the answer certainty,
The question receiving and answering system serves as a base of final answers stored in association with the respective identifiers associated with the primary answers according to the answer certainty received from the primary answer system. basic answer, or if necessary, a unique answer that is stored in association with the identification information of the user system through which the question based on the identifier and the image data has passed, and that is different in content from the basic answer, from the final answer storage unit 7. The question answering system according to any one of claims 1 or 3 to 6, or any one of claims 2 to 6, wherein the final answer to the question based on the image data is read out and transmitted to the question terminal. or the question reception and answer system described in paragraph 1. moreover,
A record of communication with the question terminal, a record of communication with the primary response system, and an operation history memory for storing various information from the reception of the question by the image data to the final answer, including unique attribute information of the question by the image data. Department and
According to the authority, from the user system, maintenance including viewing of the various information stored in the operation history storage unit and correction, addition, and change of the final answer candidate stored in the final answer storage unit a maintenance means capable of
5. The question answering system according to any one of claims 1, 3 or 4 , or the question receiving and answering system according to any one of claims 2 to 4, comprising: moreover,
A record of communication with the question terminal, a record of communication with the primary response system, and an operation history memory for storing various information from the reception of the question by the image data to the final answer, including unique attribute information of the question by the image data. Department and
According to the authority, the user system can view the various information stored in the operation history storage unit, and modify, add, and modify the final answer candidate or the unique answer stored in the final answer storage unit. a maintenance means capable of maintenance including modification;
8. The question answering system according to any one of claims 5 to 7, or the question receiving and answering system according to any one of claims 5 to 7, comprising: The contents of the final answer candidate, the basic answer, or the unique answer stored in the final answer storage unit are restricted to the assumed questions of the primary answer system by the maintenance means according to authority. 10. The question answering system according to claim 9 , or the question receiving and answering system according to claim 9, characterized in that it is possible to add, modify, change, and delete updates. A question receiving and answering system for receiving a question based on image data from a question terminal by a computer and replying an answer to the question based on the image data to the question terminal; A question in a question answering system comprising: a primary answer system that machine-learns a predetermined identifier that is an answer of as a primary answer and returns the identifier as a primary answer in response to a question based on input image data A method of responding,
(a) in the question acceptance and answer system, a step of accepting a question by image data from a question terminal and transmitting the question by the image data to the primary answer system;
(b) In the primary response system, the question based on the image data is received from the question receiving and answering system, the question based on the image data is input to the artificial intelligence, and the question based on the input image data is answered. a step of outputting the primary response as the identifier and transmitting the identifier of the primary response output from the artificial intelligence to the question receiving and answering system;
(c) In the question receiving and answering system, when the identifier is received as the primary answer from the primary answering system, it is linked to the identifier machine-learned by the artificial intelligence and the final answer candidate a step of reading the final answer candidate corresponding to the received identifier from the final answer storage unit storing the final answer candidate, and replying the read final answer candidate to the question terminal as a final answer;
A question answering method for a question answering system comprising: A question receiving and answering system for receiving a question based on image data from a question terminal by a computer and replying an answer to the question based on the image data to the question terminal; and a primary answer system that machine-learns a predetermined identifier that is the answer of as a primary answer and returns the identifier as a primary answer in response to a question based on the input image data. A question answering method for a question receiving and answering system,
(a1) in the question acceptance and answer system, a step of accepting a question based on the image data from a question terminal and transmitting the question based on the image data to the primary response system;
(b1) When the identifier is received as the primary response from the primary response system, the final response is linked to the identifier machine-learned by the artificial intelligence and stored as a final response candidate. a step of reading the final answer candidate corresponding to the received identifier from a storage unit and replying the read final answer candidate as a final answer to the question terminal;
A question answering method for a question receiving and answering system. The step (a) or (a1) comprises:
(a-1) receiving a question based on the image data from the question terminal;
(a-2) acquiring and storing unique attribute information of the question accompanying the question based on the image data;
(a-3) sending a question based on the image data to the primary response system;
with
The step (c) or (b1) is
(c-1) receiving the identifier as the primary response from the primary response system;
(c-2) obtaining identification information of the user system from the unique attribute information stored in step (a-1);
(c-3) Identifying the identifier of the primary response and the user system from the response storage unit that stores the final response candidates corresponding to the identifier, divided into basic responses and user-specific unique responses. retrieving the corresponding basic answer or unique answer based on the information;
(c-4) returning the read basic answer or unique answer to the question terminal;
The question answering method of the question answering system according to claim 10 or the question answering method of the question receiving and answering system according to claim 12 , characterized by comprising: