KR102374530B1 - Optimized question and answer system and method thereof - Google Patents

Abstract

The present invention discloses a question and answer system. More particularly, the present invention relates to an optimal question-answering system and method for deriving and replying optimal responses to various types of queries from users in a machine learning-based interactive messenger program or the like.
According to an embodiment of the present invention, answer content is extracted through a plurality of reasoners that derive intents according to different criteria for a query message, sorted by priority according to the inference rate, and provided to the user, thereby increasing the user's intention It has the effect of providing an answer that is consistent with

Description

OPTIMIZED QUESTION AND ANSWER SYSTEM AND METHOD THEREOF

The present invention relates to a question-and-answer system, and to an optimal question-answering system and method applied to a machine learning-based interactive messenger program and the like to derive and reply optimal responses to various types of queries from users.

In an interactive messenger program such as a chatter robot, when a user who wants a consultation uses a message transmission method to input a question in everyday language as if talking with a person using a message, artificial intelligence (AI) generates big data Based on the analysis, it is possible to provide an interactive messenger service that provides answers as if chatting with a person through a general language.

In accordance with the 4th industrial revolution era, these interactive messenger programs are gradually expanding their application areas not only to simple consultations such as product sales and inquiries in online shopping malls, but also to remote medical consultations that require specialized knowledge.

In general, an interactive messenger program interprets a message input from a user, automatically generates a reply message, and then provides it to the user. However, such a program has a limitation that, when an unprepared, mechanically unresponsive question is input, an inappropriate answer to the question is often provided.

Accordingly, at the current technology level, the interactive messenger program can handle a small amount of data work that can be created in a short period of time, simple consultation work, or even unskilled people without product knowledge such as branch location information or phone number information. It can only be applied to simple inquiries, and it is a trend that requires skills to build professional counseling knowledge and learn programs at the level of an agent based on this.

Registered Patent Publication No. 10-2103273 (Announcement Date: 2020.05.21.)

The present invention has been devised to overcome the above-mentioned limitations, and the present invention is applied to an interactive messenger system in which a user can perform a consultation through a message transmission means, including the medical consultation field, and can respond to various situations. It is an object of the present invention to provide a system and method for providing a more optimized response to a user through a plurality of reasoning units.

In order to solve the above problem, the optimal question answering system according to an embodiment of the present invention provides a response to one or more user terminals receiving a query message including query contents from a user, and one query message in different specific patterns. Through a content database server and information communication network that extract a plurality of different answer contents according to each other, the user terminal and the content database server are connected to each other to transmit a query message, and respond to a plurality of the query messages from the content database server It may include a relay server that receives a plurality of answer content, sorts according to the inference rate and provides the content to the user terminal.

The user terminal may access a web page posted by the relay server through a mounted web browser, receive the query message, and transmit it.

The relay server provides a web page providing unit that publishes the web page online and an input interface on the web page to receive the query message, and transmits the query message to the content database server. may include wealth.

The content database server includes an answer content database in which a plurality of answer content is stored, a plurality of different reasoners for inferring intents with respect to the query message, and inferring intents according to a specific pattern An inference unit for extracting answer content corresponding to the content from the answer content database, a learning database in which learning data in which relationships between various intents and answers are defined are stored, and a plurality of learning engines performing machine learning, It may include a learning unit that inputs the learning data to the learning engine to derive a specific pattern and provides it to the reasoning unit.

The reasoning unit may calculate an inference rate representing the degree of matching between the answer content extracted from each reasoner and the query message as a numerical value, and transmit the answer content according to the calculated inference rate to the relay server.

The result arrangement unit of the relay server determines a priority according to the inference rate for a plurality of answer contents according to the inference rate output from the plurality of reasoning units, and arranges the plurality of answer contents according to the priority in ascending or descending order can be sorted by

In addition, in order to solve the above-described problem, an optimal query response method according to another aspect of the present invention is an optimal query response method by a relay server and a content database server according to claim 1, wherein the relay server is receiving a query message including query content from a user terminal through a page and transmitting it to the content database server; It may include the step of transmitting, and the relay server arranging a plurality of answer content according to the inference rate and transmitting the content to the user terminal.

Before the content database server extracts a plurality of answer content through a plurality of different reasoners and transmits it to the relay server, the content database server inputs learning data to a plurality of learning engines that perform machine learning It may further include the step of deriving a specific pattern and providing it to the reasoning machine.

The step of extracting, by the content database server, a plurality of answer contents through a plurality of different reasoners and transmitting the content to the relay server includes the steps of: inferring, by the content database server, an intent according to a specific pattern through the reasoning machine; The method may include extracting, by the reasoner, answer content corresponding to the inferred intent from an answer database.

The extracting of the answer content corresponding to the inferred intent by the reasoner from the answer database includes the steps of: calculating, by the reasoning machine, a reasoning rate representing the degree of correspondence between the extracted answer content and the query message as a numerical value; It may include transmitting the answer content according to the determined inference rate to the result organizer of the relay server.

In the step of transmitting the answer content according to the calculated inference rate to the result organizing unit of the relay server, the result organizing unit determines a priority according to the inference rate with respect to a plurality of answer contents according to the inference rate output from the reasoning unit and arranging a plurality of answer contents in ascending or descending order according to the priority.

According to an embodiment of the present invention, answer content is extracted through a plurality of reasoners that derive intents according to different criteria for a query message, and the content is sorted by priority according to the inference rate and provided to the user, thereby increasing the user's intention. It has the effect of providing an answer that is consistent with

1 is a diagram schematically showing an overall connection structure of an optimal question and answer system according to an embodiment of the present invention.
FIG. 2 is a diagram showing the structures of a relay server and a content database server of an optimal question and answer system according to an embodiment of the present invention.
3 to 6 are diagrams illustrating an optimal question answering method according to an embodiment of the present invention.
7 is a diagram illustrating one screen of a web page for entering a query provided by the optimal question answering system according to an embodiment of the present invention.

When a part "includes" or "includes" a certain component throughout the specification, this means that other components may be further included, rather than excluding other components, unless otherwise stated. In addition, terms such as "...unit", "...terminal" and "...system" described in the specification refer to a unit that processes one or more functions or operations. Meaning, it may be implemented in hardware, software, or a combination of hardware and software.

Also, in this specification, the term "embodiment" is meant to serve as an illustration, example, or illustration, but the subject matter of the invention is not limited by such examples. Also, although "comprising", "comprising", "having" and other similar terms are used, when used in the claims, "comprising" is an open transition word that does not exclude any additional or other elements. It can be used generically in a manner similar to the term "Comprising".

The various techniques described herein may be implemented with hardware or software, or a combination of both, where appropriate. As used herein, terms such as "...Unit", "...Terminal" and "...System" likewise refer to computer-related entities, i.e., hardware , a combination of hardware and software, software or software in execution may be treated as equivalent. In addition, each function executed in the program of the present invention may be configured in module units, and may be recorded in one physical memory, or may be recorded while being dispersed between two or more memories and recording media.

In the following description, the term "optimal question and answer system" of the present invention may be abbreviated as "question answer system" or "system" for convenience of description.

1 is a diagram schematically showing an overall connection structure of an optimal question and answer system according to an embodiment of the present invention.

Referring to FIG. 1 , the optimal question answering system according to an embodiment of the present invention includes one or more user terminals 100 receiving a query message including query contents from a user, each user terminal 100 through an information communication network, and It is connected to the content database server 300 to transmit a query message, and receives a plurality of answer contents corresponding to the plurality of query messages from the content database server 300 and sorts according to the inference rate to the user terminal 100 The relay server 200 and the content database server 300 for extracting a plurality of different answer content according to different specific patterns with respect to one query message may be included.

The user terminal 100 is a terminal device possessed by a user for conducting medical consultation using the system of the present invention, and a known smart phone, a stationary PC, etc. can be used, and as a communication means is mounted, the first communication network It is connected to the relay server 200 through , and can access a web page posted online by the relay server 200 to use a messenger service.

The user terminal 100 may be equipped with a known web browser such as Internet Explorer or Chrome, and may display a search box included in a web page. The user can use the AI-based search service by entering a desired query message through the search box. In particular, the system according to an embodiment of the present invention can provide a messenger service in connection with a plurality of medical institutions, and the user can search for a medical institution through a natural language query or inquire about medical related consultation contents.

As such a user terminal 100, PCS (Personal Communication System), GSM (Global System for Mobile communications), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication) )-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet) terminal, smartphone, smartpad, tablet PC (Tablet) All kinds of handheld-based mobile terminal devices such as PC) and computing devices such as stationary PCs and notebook computers may be used.

The relay server 200 may receive a query message from a plurality of user terminals 100 connected through the first communication network and transmit it to the content database server 300 , and also respond to the query message from the content database server 300 . By receiving the corresponding answer content and replying to the corresponding user terminal 100, it is possible to implement an artificial intelligence-based service.

To this end, the relay server 200 transmits the input query message as well as the web page providing means to the content database server 300 connected through the second communication network, and inferred a plurality of query messages provided therefrom. It may include a result delivery means for arranging the answer content of the according to the priority and delivering it to the user terminal 100 . A detailed description of the structure and functions of the relay server 200 will be described later.

The relay server 200 may be implemented as a well-known web server such as Apache, ISS, nginx, etc. that can publish a web page composed of prepared HTML, Java script, etc. online.

As the first and second communication networks described above, long-term evolution (LTE), LTE Advance (LTE-A), code division multiple access (CDMA), wideband CDMA (WCDMA), universal mobile telecommunications system (UMTS), WiBro Any one of (Wireless Broadband) or Global System for Mobile Communications (GSM) may be used.

The content database server 300 includes a reasoner derived through artificial intelligence-based learning, and infers a plurality of intents for the query message through a plurality of reasoners using a specific pattern derived by the learning engine. and extracting one or more of the plurality of answer contents prepared in advance according to the correlation with the intents to extract a plurality of answer contents according to different specific patterns.

The content database server 300 may be implemented as a well-known relational or object type database system, such as IBM DB2, Oracle, MySQL, or MS-SQL server.

According to the above-described structure, the optimal question answering system according to an embodiment of the present invention provides a system when a user inputs an inquiry through a query message composed of natural language through his/her user terminal online during consultation with a medical institution, etc. can provide an optimal response that best meets the user's intentions by independently extracting a plurality of answer contents to a single query message through a plurality of different reasoners, sorting them according to priority, and replying to the user terminal.

Hereinafter, the relay server and the content database server 300 of the optimal question and answer system according to an embodiment of the present invention will be described in detail with reference to the drawings.

FIG. 2 is a diagram showing the structures of a relay server and a content database server of an optimal question and answer system according to an embodiment of the present invention. In the following description, the constituent units of each server can be implemented by a known microprocessor, recorded on a readable and writable recording medium, and mounted on the server device.

Referring to FIG. 2 , the optimal question and answer system according to an embodiment of the present invention provides a web page and performs machine learning and a relay server 200 that connects the user terminal 100 and the content database server 300 to each other. It may include a content database server 300 that finds a specific pattern existing in natural language in a query message through the .

In detail, the relay server 200 provides a web page providing unit 210 that publishes a web page online, an input interface on the web page to receive a query message, and transmits the input query message to the content database server. It may include a query unit 220 that transmits to 300 .

The web page providing unit 210 may publish a web page written in a known language such as HTML or CSS online. The web page may provide a search interface, and the user terminal 100 may access the web page through a web browser and input a query message through the search interface.

The query unit 220 may transmit a query message input through the web page of the web page providing unit 210 to the content database server 300 . In this case, the query message may be text that is composed of a plurality of characters such as Korean or English to form one sentence, and the query message is input to a plurality of reasoning units provided in the content database server 300 , respectively. This allows multiple inference results to be returned.

The result organizing unit 230 may receive the extracted search content for the query message transmitted from the content database server 300 and reply to the user terminal 100 having input the query message.

In particular, the result organizing unit 230 transmits all of the plurality of answer contents extracted by the reasoning unit 320 to the user terminal 100 , but a plurality of answer contents according to the inference rate output from the plurality of reasoning units 325 . , a priority may be determined according to an inference rate, and a plurality of answer contents may be sorted in ascending or descending order according to the priority. Accordingly, the user may sequentially check one or more answer contents searched for as a result of a query through the user terminal 100 according to their ranking, and may obtain intended information.

According to the above-described structure, the relay server 200 may provide the query message of the user terminal 100 to the content database server 300 and return the answer content.

In addition, the content database server 300 includes an answer content database 310 in which a plurality of answer content is stored, and a plurality of different reasoners 325 for inferring an intent with respect to a query message, It includes an inference unit 320 that infers an intent according to a pattern and extracts answer content corresponding to the inferred intent from the answer database 310, a plurality of learning engines that perform machine learning, and learns the learning data. It may include a learning unit 330 input to the engine 335 to derive a specific pattern and provided to the reasoning unit, and a learning database 340 in which learning data in which relationships between various intents and answers are defined are stored.

The answer content database 310 may store a plurality of answer content for a plurality of intents, and may provide answer content corresponding thereto according to a specific pattern of the reasoner 325 . Here, each answer content database 310 may store answer data that is updated as different specific patterns learned by different learning engines are provided to the reasoning unit 325 and also different data.

The inference unit 320 may infer an intent based on the query message and extract answer content according to the intent. The reasoning unit 320 according to an embodiment of the present invention may include a plurality of reasoners 325 derived by a plurality of different learning engines, and analyzes a query message to infer an intent thereof.

As an example, if the query message is a single sentence, "Tell me the type of LASIK surgery," this sentence is a plural of "LASIK" + " "+ "Surgery" + " " + "Type" + " " + "Tell me". It can be parsed into words or tokens, and the reasoning unit 325 derives a pattern of a sentence according to the meaning and word order of each of the parsed words or tokens, and inputs the sentence to a specific pattern derived by the learning engine 335 . What this means, 'intents', is inferred.

In particular, the reasoner 325 may derive intents having different results by analyzing the query message using a specific pattern by the learning engine 335 that has performed machine learning with different learning data. . In addition, the reasoner 325 may extract at least one most suitable answer content from among a plurality of answer content stored in the answer content database 310 .

In addition, the reasoning unit 320 calculates an inference rate representing the degree of matching between the extracted answer content and the query message as a numerical value, and transmits the answer content according to the calculated reasoning rate to the result organizing unit 230 of the relay server 200 . can be transmitted Accordingly, the result organizing unit 230 may receive a plurality of answer content having an inference rate extracted by the plurality of reasoning units 325, and determine the priority of the plurality of answer content according to the inference rate, and prioritize The plurality of answer contents are arranged according to the order so that they are sequentially displayed on the web page.

The learning unit 330 may derive the above-described specific pattern using different supervised learning techniques or different learning data, and may independently provide the above-described specific patterns to the reasoning unit 325 .

The learning unit 330 may include a plurality of learning engines 335 , and the learning engine 335 uses a variety of learning data stored in the learning database 340 through machine learning to specify specific words, word order, and sentences. A pattern may be derived and provided to the reasoner 325 . As an example, "eye", "myopia", and "ophthalmology" may be derived from "LASIK", and "treatment", "operation" and the like may be derived from "surgery".

The learning database 340 may store a plurality of learning data. The learning data may improve the accuracy of the learning engine 335 by providing a plurality of learning data stored during machine learning by the learning engine 335 .

In particular, as the system of the present invention is a system specialized for a medical institution, it can be implemented to hold contents specialized for hospital work by building specialized terms or terms with high frequency used in medical institutions such as hospitals and public health centers as learning data.

Hereinafter, an optimal method for answering a question by a system according to an embodiment of the present invention will be described with reference to the drawings.

3 to 6 are diagrams illustrating an optimal question answering method according to an embodiment of the present invention. Each step in the description below

First, referring to FIG. 3 , the optimal query response method according to an embodiment of the present invention is an optimal query response method using a relay server and a content database server. receiving a message and transmitting it to the content database server (S100), extracting a plurality of answer content through a plurality of different reasoners by the content database server and transmitting the content to the relay server (S200), and the relay server It may include a step (S300) of arranging a plurality of answer content according to the inference rate and transmitting it to the user terminal.

First, the relay server receives a query message including the query content from the user terminal through the web page and transmits it to the content database server (S100). A user who wants to use the messenger service of the present invention uses the user terminal Thus, when the web page of the present invention is accessed and a query message is input in the search box, the relay server receives and transmits the query message to the content database server.

Next, the content database server extracts a plurality of answer contents through a plurality of different reasoners and transmits them to the relay server (S200). The reasoning unit parses the input query message to analyze the sentence, and for each reasoner The intent is inferred according to a specific pattern derived as a result of machine learning of the learning engine based on different learning data, and the answer content corresponding to the intent is extracted from the answer content database.

Next, in the step (S300) of the relay server arranging the plurality of answer contents according to the inference rate and transmitting it to the user terminal, the content database server transmits the plurality of answer contents extracted in the step S200 to the relay server, as well as reasoning The priority is determined based on the inference rate for each answer content provided by the supplement, and the order is arranged according to the order and transmitted to the user terminal.

In addition, referring to FIG. 4 , in step S200 described above, the content database server infers the intent according to a specific pattern through the reasoner ( S210 ) and the reasoning unit returns the answer content corresponding to the inferred intent to the answer data It can be subdivided into the step of extracting from the base (S220).

In detail, as a step (S210) of the content database server inferring the intent according to a specific pattern through the reasoning unit, the inference unit of the content database server receives the intent for the query message according to a plurality of different specific patterns. The reasoning machine for inference is included, and the intent is inferred according to different criteria for each reasoner.

Next, in the step of extracting the answer content corresponding to the inferred intent by the reasoner from the answer database (S220), extracting the answer content corresponding to the inferred intent for each reasoner from the answer database, As one or more answer contents are extracted for each group, a plurality of answer contents may be extracted with respect to one query message.

In addition, referring to FIG. 5 , in step S220 described above, the reasoning unit calculates an inference rate representing the degree of agreement between the query message and the answer content as a numerical value ( S221 ) and relaying the answer content according to the calculated inference rate to the relay server It may include a step (S222) of transmitting the result to the organizing unit.

In the step (S221) of the reasoning machine calculating the inference rate representing the degree of correspondence between the inferred intent and the extracted answer content as a numerical value, in extracting the answer content using the intents of each reasoner, the extracted answer content and The degree of matching of the query message, which is the raw data, can be calculated as a percentage numerical value.

Next, the step of transmitting the answer content according to the calculated inference rate to the result organizing unit of the relay server (S222) is to transmit the plurality of answer content extracted by the reasoning machine and the inference rate to the result organizing unit, so that the user through the web page after organizing to be transmitted to the terminal.

And, referring to FIG. 6, the above-described step S222 is a step (S2221) of determining a priority according to the inference rate with respect to the plurality of answer contents according to the inference rate output from the reasoning unit by the result organizing unit (S2221) and a plurality of pieces according to the priority It may include arranging the answer content of the in ascending or descending order (S2222).

In detail, the step (S2221) of determining the priority according to the inference rate with respect to the plurality of answer contents according to the inference rate output from the reasoning unit by the result organizing unit according to the inference rate of the answer content provided from the content database server The result organizer of the relay server determines the ranking of the answer content according to the numerical value or percentage.

Next, in the step of arranging the plurality of answer contents in ascending or descending order according to the priority (S2222), by sequentially arranging the plurality of answer contents according to the determined priority and reflecting it on the web page, the user accessing the web page to check the answer content in the order of highest accuracy.

Hereinafter, the technical idea of the present invention will be described in detail through an example of a screen of a web page provided by a system according to an embodiment of the present invention.

7 is a diagram illustrating one screen of a web page for entering a query provided by the optimal question answering system according to an embodiment of the present invention.

Referring to FIG. 7 , the optimal question-and-answer system according to an embodiment of the present invention may provide a homepage or web page for requesting a visit reservation, medical inquiry, etc. of customers to a medical institution, etc., and such a web page W100 is It may include a search window W110 in which a plurality of users connected through their user terminal can input a query message.

As an example, the user may input a query message “Tell me the type of LASIK surgery” into the search window W110, and the relay server transmits the entered query message to the content database server to infer the intent and receive the answer content. Thus, it can be displayed through the web page W100.

In addition, the relay server may display information on the medical institution currently registered in the system and the medical staff belonging to the institution through the web page W100. In particular, the notification window W120 for this purpose is the content of the inquiry according to the received answer content. Can indicate the number of medical institutions or medical staff that can be connected to

In addition, the web page W100 may include a result window W130 displaying a plurality of answer contents provided by the content database server. In the result window W130, a priority set based on an inference rate for a plurality of answer contents is applied, and a plurality of answer contents may be sequentially displayed in ascending or descending order according to the priority. 7 shows an example of displaying information on a medical institution registered in the system as response content to a query message.

The user confirms the information of the medical institution included in the above-mentioned result window W130, selects any one medical institution according to the intention, and then continues the conversation with the interactive messenger, or directly chats with the medical institution can also be done

Although many matters are specifically described in the above description, these should be construed as examples of preferred embodiments rather than limiting the scope of the invention. Accordingly, the invention should not be defined by the described embodiments, but should be defined by the claims and equivalents to the claims.

100: user terminal 200: relay server
210: web page providing unit 220: query unit
230: result organizer 300: content database server
310: answer content database 320: reasoning unit
325: reasoning machine 330: learning unit
335: learning engine 340: learning database

Claims (11)

one or more user terminals for receiving a query message including query contents from a user;
a content database server for extracting a plurality of different answer contents according to different specific patterns with respect to one query message; and
Each is connected to the user terminal and the content database server through an information communication network to transmit a query message, and receives a plurality of answer contents corresponding to the query message from the content database server and sorts according to the inference rate, the user terminal including a relay server provided to
The content database server,
an answer content database in which a plurality of answer content is stored;
a plurality of different reasoners for inferring intents with respect to one query message; an inference unit extracting from the content database;
a learning database in which learning data in which relationships between various intents and answers are defined; and
It includes a plurality of learning engines to which different techniques for performing machine learning are applied, and a learning unit that inputs the learning data to the learning engine to derive a specific pattern, and provides it to a plurality of reasoning machines,
The reasoning unit is
An optimal question answering system that calculates an inference rate representing the degree of matching between a plurality of answer content extracted from each reasoner and the query message as a numerical value, and transmits the answer content according to the calculated inference rate to the relay server . The method of claim 1,
The user terminal is
An optimal question-and-answer system for receiving and transmitting the query message by accessing a web page posted by the relay server through a built-in web browser. 3. The method of claim 2,
The relay server is
a web page providing unit for posting the web page online; and
A query unit that provides an input interface on the web page, receives the query message, and delivers the input query message to the content database server
An optimal question-and-answer system that includes. delete delete The method of claim 1,
The result organizing unit of the relay server,
An optimal question-and-answer system for determining a priority according to the inference rate with respect to a plurality of answer contents according to the inference rate output from the plurality of reasoners, and arranging the plurality of answer contents in ascending or descending order according to the priority. An optimal method for answering questions by the relay server and content database server according to claim 1,
receiving, by the relay server, a query message including query contents from a user terminal through a web page, and transmitting the received query message to the content database server;
inputting, by the content database server, learning data to a plurality of learning engines that perform machine learning to which different techniques are applied, deriving a specific pattern, and providing it to a plurality of reasoning units;
extracting, by the content database server, a plurality of answer content corresponding to the query message through a plurality of different reasoners and transmitting the content to the relay server; and
Comprising the step of the relay server arranging a plurality of answer content according to the inference rate and transmitting it to the user terminal,
The step of extracting, by the content database server, a plurality of answer content corresponding to the query message through a plurality of different reasoners and transmitting the content to the relay server,
calculating, by a plurality of reasoning units, intents according to a specific pattern, respectively, and calculating an inference rate indicating a degree of correspondence between a plurality of answer contents corresponding to the plurality of inferred intents and the degree of correspondence between the query message as a numerical value; and
and transmitting a plurality of answer contents according to the calculated inference rate to a result organizing unit of the relay server. delete delete delete 8. The method of claim 7,
Transmitting the answer content according to the calculated inference rate to the result organizing unit of the relay server comprises:
determining, by the result arrangement unit, a priority according to an inference rate with respect to a plurality of answer contents according to the inference rate output from the reasoning unit; and
Sorting a plurality of answer contents in ascending or descending order according to the priority
An optimal question-and-answer method including

Images