KR20190066988A - Chatbot system based on artificial intelligence markup language and method for proding chatbot service - Google Patents

Abstract

Disclosed are an AIML-based chatbot system and a chatbot service providing method thereof. According to the present invention, the AIML-based chatbot system comprises a machine learning server and a chatbot server. The machine learning server generates a learning model for detecting an intention corresponding to a question by machine-learning a plurality of data sets which are a data set including a question and an intention corresponding thereto, and detecting the intention corresponding to the question by using the generated learning model. Also, the chatbot server stores a plurality of intention patterns defined in an AIML form and templates corresponding to the intention patterns to respond to the question of a user, recognizes the intention corresponding to a query through the machine learning server for the query received from a user terminal, and searches for the intention patterns matching the recognized intention to provide the user terminal with the templates corresponding to the searched intention patterns as a response for the query.

Description

Technical Field [0001] The present invention relates to an AIML-based chatbot system and a method of providing a chatbot service,

The present invention relates to an AIML-based chatbot system and a method for providing a chatbot service.

Recently, as a result of expanding the usage environment of messenger via smartphone and development of artificial intelligence, customers and companies want to be connected through messaging service. Therefore, chatbot (messenger) ChatBot) system is in the spotlight. Such a chatbot system is originally a robot that communicates with a system that means a chatter robot system.

The chatbot system is programmed using Artificial Intelligence Markup Language (AIML), a non-standard generation language based on XML (Extensible Markup Language).

In general, AIML-based methods for natural language processing and understanding are defined by using expressions for pattern matching defined in the AIML specification. An example of natural language pattern matching definition using AIML is as follows.

<category>

<pattern> WHAT IS YOUR </ pattern>

<template> My name is Michael NS Evanious. </ template>

</ category>

 In the pattern matching definition, the contents of the <pattern /> tag are pattern expressions for processing / understanding of the natural language entered by the user. The "*" as defined here is a wildcard defined in AIML and is used to match one or more "random" words or phrases. In detail, all of the sentences beginning with "WHAT IS YOUR" ("WHAT IS YOUR NAME", "WHAT IS YOUR ID", etc.) are matched and the contents of the <template />, "My name is Michael NS Evanious." Is a dialogic knowledge representation for the "question-and-answer" where the response is output to the user.

In order to develop a chatbot system that understands natural language, various sentence examples are collected. In order to process and understand the natural language of AIML based on these sentences, Should be created.

If the range of the chatbot service is narrow and the type of the natural language query is not varied, it can be easily expressed by the pattern expression of AIML. However, if the range of the service is wide and the type of the natural language question is long and complicated, it is easy to define the flexible pattern expression Therefore, natural language processing and understanding performance deteriorates, and there is a problem that not only initial knowledge construction but also substantial operation and maintenance after installation cost considerable expense and effort.

A problem to be solved by the present invention is to understand a question of a natural language based customer through a machine learning based system and to execute a scenario defined in AIML based on the AIML-based chatbot server by utilizing the system intrinsic value of a question intention and a reliability value And a method of providing a chattb service using the AIML-based chatbot system.

In a chatbot system according to one aspect of the present invention,

A chatbot system based on Artificial Intelligence Markup Language (AIML), comprising: a machine learning module for generating a learning model for detecting intentions corresponding to a question by mechanically learning a plurality of data sets as a data set composed of questions and intentions corresponding to the questions; A machine learning server for detecting an intention corresponding to a question using a learned learning model; Storing a plurality of intention patterns defined in an AIML form and a template corresponding to the intention pattern by responding to a user's query, and storing an intention corresponding to the query term through the machine learning server for a query term received from the user terminal And a chat server for searching for an intention pattern matched to the intention and providing a template corresponding to the retrieved intention pattern to the user terminal in response to the query term.

Here, the chatbot server may include an intention-based database for storing the plurality of intention patterns and templates corresponding to the intention patterns; A pattern matching unit for searching for a pattern matching the intention identified through the machine learning server; And a template extracting unit for extracting a template corresponding to the pattern retrieved by the pattern matching unit.

In addition, the pattern matching unit performs pattern matching on the query word through the intention-based database, and when the matched pattern is not searched, the pattern matching unit transmits the query word to the machine learning server, And intention corresponding to the query term is grasped by receiving the intention.

In addition, the machine learning server may assign a reliability value for the intention of the machine learning to the data set and store the reliability value in correspondence with the learned intention in the learning model, and when the query term is received from the pattern matching unit, And provides the pattern matching unit with the intention and the confidence value corresponding to the query term.

The pattern matching unit may perform pattern matching using the intent received from the machine learning server when the reliability value received from the machine learning server is equal to or greater than a preset threshold value, And provides a response detection failure to the user terminal in response to the query.

In addition, a plurality of intention patterns stored in the intention-based database are defined only as words or sentences without using wildcards.

According to another aspect of the present invention, there is provided a method of providing a chat service,

A method for providing a chattb service to a user terminal includes an intention-based database for storing a plurality of intention patterns defined in an AIML form by responding to a user's query and a template corresponding to the intention pattern, Receiving a query term from the user terminal; Receiving a query from the machine learning server after transmitting the query to a machine learning server, the machine learning server comprising a machine learning module for learning a plurality of sets of data as a data set comprising a question and an intention corresponding to the question, Generating a learning model for detecting the intention, and using the generated learning model to detect intent corresponding to the question; Retrieving a pattern matching the intention received from the machine learning server using the intention-based database; And extracting a template corresponding to the retrieved pattern from the intention-based data face and providing the retrieved template as a response corresponding to the query term to the user terminal.

Searching for a pattern matched using the intention-based database for the query term, between receiving the query term and receiving the intent; And providing a response corresponding to the query term if a pattern matching the query term is retrieved from the intention based database, or if the pattern matching the query term is not retrieved from the intention based database, And performing a receiving step.

In addition, in the step of receiving the intention, further receiving a reliability value corresponding to the intention from the machine learning server, and before the step of searching for the pattern, determining whether the reliability value is equal to or greater than a preset threshold value; And searching the pattern if the confidence value is greater than or equal to the preset threshold value or performing a step of providing a response detection failure to the user terminal in response to the query if the confidence value is less than the preset threshold value .

In a chatbot system according to another aspect of the present invention,

A chattbot system based on AIML (Artificial Intelligence Markup Language), comprising: an intention-based database storing a plurality of intention patterns defined in AIML form by responding to a user's query; and a template corresponding to the intention pattern; A plurality of data sets are transmitted to an external cloud server as a data set composed of a question and an intention corresponding to the question to perform machine learning for intent detection corresponding to a question, A pattern matching unit for delivering the pattern to the cloud server to receive an intention corresponding to the query term and searching for a pattern matching the received intention, the external cloud server being a server capable of machine learning; And a template extracting unit for extracting a template corresponding to the pattern retrieved by the pattern matching unit.

Here, the pattern matching unit performs pattern matching on the query word through the intention-based database, and when the matched pattern is not found, the pattern matching unit transmits the query word to the external cloud server, And receives the corresponding intent.

The pattern matching unit further receives the reliability value for the intention from the external cloud server, and performs pattern matching using the intention received from the external cloud server when the reliability value is equal to or greater than a preset threshold value.

According to another aspect of the present invention,

A communicative system based on Artificial Intelligence Markup Language (AIML), comprising a communicator, a memory and a processor, the communicator comprising: a user terminal and a machine learning server, the machine learning server comprising a data set comprising a question and an intention corresponding to the question And a server for generating a learning model for detecting an intention corresponding to a question and detecting the intention corresponding to the question using the generated learning model, The method comprising: storing a plurality of intention patterns defined in AIML form and a template corresponding to a plurality of intent patterns for responding to a user's query and storing a set of codes, To the machine learning server via the communicator .; Receiving an intention provided from the machine learning server via the communicator; Retrieving from the memory an intent pattern that is matched against the intent; Extracting a template corresponding to the retrieved intent pattern from the memory; And providing the extracted template to the user terminal via the communicator.

Wherein the processor is further configured to: search in the memory for a pattern that is matched against the query term; And transmitting the query to the machine learning server via the communicator if a pattern matching the query is not found.

The processor may further include: an operation of further receiving a reliability value corresponding to an intention received from the machine learning server; Determining whether the reliability value is greater than or equal to a preset threshold value; And performing pattern matching on the intention received from the machine learning server when the reliability value is equal to or greater than the preset threshold value.

According to an embodiment of the present invention, a machine learning-based system is used to understand a question of a natural language based customer and to execute a scenario defined in the AIML-based chatbot server using a question intention and a reliability value, It can provide a way to be able to.

In addition, it is possible to improve the processing and understanding of natural language processing and improve the development and operation efficiency by learning and automating the learning through the machine learning system without manually defining the corresponding pattern expression in AIML for the function of understanding the natural language.

In addition, realization of continuous conversation, context management, and scenario, which are the advantages of AIML, can be provided and a more complete chatbot service can be provided through interworking with a machine learning system using a large learning DB.

1 is a schematic block diagram of a chatbot system according to an embodiment of the present invention.
FIG. 2 is a detailed block diagram of the chatbot server shown in FIG. 1. FIG.
3 is a diagram showing a specific configuration of the machine learning server shown in FIG.
FIG. 4 is a diagram showing an example of a question-intention data set used by the machine learning server shown in FIG. 1 for learning.
5 is a schematic flowchart of a method of providing a chat service in accordance with an embodiment of the present invention.
6 is a diagram illustrating a screen displayed on a user terminal according to a method of providing a chat service in accordance with an embodiment of the present invention.
FIG. 7 is a diagram showing an example of an intention pattern stored in the intention-based database shown in FIG.
FIG. 8 is a diagram showing another example of an intention pattern stored in the intention-based database shown in FIG. 2. FIG.
9 is a schematic block diagram of a chatbot server according to another embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily carry out the present invention. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In order to clearly explain the present invention, parts not related to the description are omitted, and like parts are denoted by similar reference numerals throughout the specification

Throughout the specification, when an element is referred to as "comprising ", it means that it can include other elements as well, without excluding other elements unless specifically stated otherwise.

Also, the terms " part, "" module," and " module ", etc. in the specification mean a unit for processing at least one function or operation and may be implemented by hardware or software or a combination of hardware and software have.

Hereinafter, a chatbot system according to an embodiment of the present invention will be described with reference to the drawings.

1 is a schematic block diagram of a chatbot system according to an embodiment of the present invention.

1, a chatbot system 10 according to an embodiment of the present invention includes a user terminal 100, a chatbot server 200, and a machine learning server 300. 1, the present invention is not limited to FIG. 1, and the present invention is not limited to FIG. 1, and may be configured differently from FIG. 1 according to various embodiments of the present invention. .

The user terminal 100 is connected via a network 400. In this case, the network mainly refers to a connection structure capable of wirelessly exchanging information between respective nodes such as terminals and servers as a wireless network. Examples of such a network include a 3rd Generation Partnership Project (3GPP), an LTE Long Term Evolution, World Interoperability for Microwave Access (WIMAX), Wi-Fi, 3G, 4G, and 5G.

The user terminal 100 receives a query of the user and transmits the query to the chat server 200 via the network 400. The chat client 200 receives a response to the query from the chat server 200, Or the like.

The user terminal 100 may comprise a mobile terminal capable of wireless communication, and according to various embodiments of the present invention, the user terminal 100 may be various types of devices. For example, the user terminal 100 may be a portable terminal capable of accessing the remote chatbot server 200 through the network 400. As an example of the portable terminal, there is a wireless communication device with guaranteed portability and mobility. The wireless communication device includes a 3G, 4G, 5G terminal, an LTE terminal, a smartphone, a smart pad, a tablet PC, a personal communication system (PCS) 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) (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet) terminals, and the like. One example of the user terminal 100 is a notebook, a netbook, an UltraBook, a subnotebook, a desk note, an Ultra-Mobile PC (UMPC), a desktop ), And the like. Here, the user terminal 100 is not limited to the form shown in FIG. 1 or the ones illustrated above.

The chat server 200 searches for a pattern matching a natural language query received from the user terminal 100 and transmits a template corresponding to the retrieved pattern to the user terminal 100 in response to the query if the matching pattern is searched .

However, if the chatbot server 200 can not retrieve a pattern matching the natural language query received from the user terminal 100, the natural language query is transmitted to the machine learning server 300, A pattern corresponding to the intention can be retrieved by receiving the intention corresponding to the query term, and the template corresponding to the retrieved pattern can be transmitted to the user terminal 100 in response to the query term.

The chatbot server 200 constructs a plurality of scenarios based on the intention pattern mapping the intent pattern and the template corresponding to the intent pattern. The chatbot server 200 can provide a response to the query term of the user terminal 100 based on the plurality of scenarios based on the intention thus constructed.

The machine learning server 300 has a function of generating a corresponding learning model by mechanically learning a large amount of data as a data set composed of "question-intention ", and grasping intention of the natural language question using the generated learning model . Here, machine learning is a field of artificial intelligence. It is a technology that predicts the future by analyzing vast amounts of data. It is a technology that solves problems by acquiring information that has not been inputted by the computer through the learning process itself. Deep learning techniques using neural networks such as CNN (Convolutional Neural Network), R-CNN (Region with Convolutional Neural Network) and LSTM (Long Short Term Memory) can be used for machine learning. Since it is already well known, a detailed explanation is omitted here.

Hereinafter, the operation of each component of the chatbot system 10 shown in FIG. 1 will be described in more detail.

2 is a detailed block diagram of the chatbot server 200 shown in FIG.

2, the chatbot server 200 includes a query receiving unit 210, an intention-based database (DB) 220, a pattern matching unit 230, a template extracting unit 240, a query transferring unit 250, An intention receiving unit 260, and a response providing unit 270. Here, the chatbot server 200 shown in FIG. 2 is only one embodiment of the present invention, and various modifications are possible based on the components shown in FIG.

The query term receiving unit 210 receives a natural language query from the user terminal 100 through the network 400. [

The intention-based DB 220 stores a plurality of intention patterns defined in AIML form and a template corresponding to a plurality of intent patterns in order to respond to a user's question. Herein, the intention pattern means that a plurality of pattern expressions matched to a query word are not defined manually by AIML, but a pattern expression is simply defined based on an intention to represent a query word. In other words, conventionally, a plurality of pattern expressions corresponding to each of a plurality of query terms corresponding to a similar intention should be defined. However, an intention-based pattern according to an embodiment of the present invention may be defined as a pattern Because expressions can be defined, the definition of pattern expressions is simple and straightforward, which saves the cost of initial knowledge construction as well as operation and maintenance after construction.

The pattern matching unit 230 searches for a pattern matching a query input through the query input unit 210 from among a plurality of patterns including a plurality of intention expressions set in the intention based DB 220 in advance.

The pattern matching unit 230 transmits the retrieved pattern to the template extracting unit 240 when a pattern matching the query term is searched.

However, the pattern matching unit 230 transmits the query input from the query receiving unit 210 to the query transmission unit 250 when a pattern matching the query is not found.

The pattern matching unit 230 searches for a pattern matching the intention input through the intention receiving unit 260 among a plurality of patterns formed by a plurality of intention expressions preset in the intention based DB 220.

The template extracting unit 240 searches the intention-based DB 220 for a template corresponding to the pattern retrieved by the pattern matching unit 230 and extracts the template in response to the query.

The query language transfer unit 250 transfers the query sent from the pattern matching unit 230 to the machine learning server 300.

The intention receiving unit 260 receives the intention and the reliability value provided from the machine learning server 300 and transmits the intention and the reliability value to the pattern matching unit 230. Here, the reliability value is a confidence value corresponding to the intention learned in the machine learning server 300, and the pattern matching unit 230 only transmits the intention to the intention receiving unit 260 when the reliability value transmitted from the intention receiving unit 260 is equal to or greater than the threshold value. Is used for pattern matching. Here, the threshold value may be preset to a reliability value of 70% or the like.

The response providing unit 270 receives the response extracted by the template extracting unit 240 and provides the response to the user terminal 100 through the network 400. [

Those skilled in the art will appreciate that the query receiving unit 210, the intention based DB 220, the pattern matching unit 230, the template extracting unit 240, the query language delivering unit 250, the intention receiving unit 260, ) May be separately implemented, or one or more of them may be integrated.

Hereinafter, the machine learning server 300 shown in FIG. 1 will be described.

FIG. 3 is a diagram showing a specific configuration of the machine learning server 300 shown in FIG.

3, the machine learning server 300 includes an intention learning unit 310, a learning model database (DB) 320, and an intention detection unit 330. Here, the machine learning server 300 shown in FIG. 3 is only one embodiment of the present invention, and various modifications are possible based on the components shown in FIG.

The intention learning unit 310 generates a learning model corresponding to the learning result by performing machine learning for intention detection using a large amount of data set composed of "question-intention" input from the chattob server 200. [ Here, the question is a question corresponding to a natural language query, and machine learning is performed so that the intention is classified according to the intention pattern used in the intention-based DB 230 of the chatbot server 200. The intention learned by the intention learning unit 310 is also given a reliability value of the learning result for the intention.

An example of a "question-intent" data set, which is learning data used for machine learning in the intention learning unit 310, is shown in FIG. Referring to FIG. 4, it can be seen that the machine learning is performed such that the intention corresponding to a plurality of questions such as "let me know the charge for this month"

The learning model DB 320 stores a learning model generated by the intention learning unit 310.

The intention detection unit 330 detects the intention and the reliability value corresponding to the natural language query using the learning model stored in the learning model DB 320 for the natural language query input from the chattob server 200 and provides the intention and the reliability value to the chattob server 200 do.

Meanwhile, the intention learning unit 310 and the intention detection unit 330 may be integrated and implemented as one component. In this case, the intention learning unit 310 and the intention detection unit 330 may be implemented using one neural network.

Hereinafter, a method for providing a chat service in accordance with an embodiment of the present invention will be described with reference to the drawings.

5 is a schematic flowchart of a method of providing a chat service in accordance with an embodiment of the present invention. The chat service providing method shown in FIG. 5 can be performed by the chat server 200. [

Referring to FIG. 5, the chatbot server 200 receives a natural language query from the user terminal 100 through the network 400 (S100).

Thereafter, a pattern matching the received query is retrieved using the intention-based DB 220 (S110).

If the pattern matching the query term is not found (S120), the natural language query is transmitted to the machine learning server 300 (S130).

Thereafter, the intention and reliability value provided from the machine learning server 300 are received (S140), and it is determined whether the received reliability value is equal to or greater than a preset threshold value (S150).

If the confidence value is greater than or equal to a preset threshold value, a matching pattern is retrieved using the intention-based DB 220 for the received intention (S160).

If a pattern matching the intention is searched (S170), a template corresponding to the searched pattern is extracted from the intention-based DB 220 (S180), and the network 400 is extracted as a response to the query received in the step S100. To the user terminal 100 (S190).

However, if the reliability value is smaller than the predetermined threshold value in the step S150, the intention received from the machine learning server 300 can not be relied upon, so that the response to the query term received from the user terminal 100 has failed To the user terminal 100 (S200).

If a pattern matching the query term is found in step S120, the template corresponding to the retrieved pattern is extracted from the intention-based DB 220, and the extracted template is stored as a response to the query on the network 400 And performs the steps S180 and S190 to provide the terminal 100 with information.

Hereinafter, a chat service according to an embodiment of the present invention will be described with reference to an example.

First, as shown in FIG. 6, an example of a case where a user directly inputs a question via a keyboard through a user terminal 100 is used. That is, the question entered by the user using the user terminal 100 is "Monthly Rate" 500.

When the user inputs a question as described above, the user terminal 100 transmits the inputted question, that is, "the present month rate" as a natural language query to the chatbot server 200.

When the chatbot server 200 receives the natural language query "current month charge" transmitted from the user terminal 100, it performs pattern matching on the query term "current month charge" to search for a pattern matched from the intention based DB 230 do.

Since the chatbot system 10 according to the embodiment of the present invention works in cooperation with the machine learning server 300, the chatbot system 10 according to the embodiment of the present invention can be configured to use the pattern / > Can be defined based on the exact "intention" of the user in the form of exact words and phrases without the use of a wildcard, "*". For example, as shown in FIGS. 7 and 8, knowledge is constructed in the form of an intent-based pattern in which no wildcards are used.

Since the query term received from the user terminal 100 at the chat terminal 200, that is, the "current month charge" is not intended as an intention, the pattern matching result by the query term will not match the pattern.

Accordingly, the chattob server 200 transmits the query term "current month charge" to the machine learning server 300 and requests an intention corresponding to the query term "current month charge ". In this example, explanation is made on the assumption of "use charge inquiry" as an intention learned corresponding to the query term "current month charge ".

Therefore, the machine learning server 200 will detect "use charge inquiry" for the query term "current month charge " inputted from the chattob server 200 as an intention and provide it to the chattob server 200.

Thereafter, the chatbot server 200 receives the intention corresponding to the query term " monthly charge "provided from the machine learning server 300, that is, the" 230). &Lt; / RTI &gt;

As described above with reference to FIG. 7, the above-mentioned "usage charge inquiry" is already stored in the intention-based DB 230 as a pattern.

Accordingly, the chatbot server 200 can retrieve a pattern matching with the intention "use charge inquiry" through this pattern matching, and can search the template corresponding to the retrieved pattern, that is, The device ... "and provide it to the user terminal 100 as a response to the query term" charge for the current month ".

The user terminal 100 displays a response provided from the chat server 200 as a response 510 to the query term " current month charge " as shown in FIG.

On the other hand, the user input on the user terminal 100 can be entered directly using the keyboard, or the user can click on the keyboard to enter the content displayed as the response 510 to the " It is also possible to input in the form of a selection button.

Therefore, if the user selects, for example, "Mobile usage charge inquiry" among the displayed response 510 regarding the question "Charge for the current month ", the thus selected" (520) displayed on the user terminal 100, and the "mobile usage fee inquiry" is transmitted to the chatbot server 200 as the "monthly fee" question described above.

Thereafter, the chatbot server 200 performs the same process as described above using the "This Month Charge " However, since the response provided for the "current month rate" is a response configured by the chatbot server 200, the query included in the response may be generated based on the intention used in the chatbot server 200. [ Therefore, the second query term "mobile usage charge inquiry" may correspond to the intention of the query term itself, and thus the chatbot server 200 may perform the machine learning It is possible to retrieve a pattern directly matching through the intention based DB 230 without going through the server 300. [ It has been described above that a pattern corresponding to the intention of "inquiry of mobile usage charge" is formed in the intention-based DB 230 with reference to FIG.

Accordingly, the chatbot server 200 extracts a template corresponding to the "mobile usage charge inquiry" pattern, i.e., "... used by the customer" to the user terminal 100 as a response to the query " , And the user terminal 100 displays a response provided from the chat server 200 as a response 530 to the query "Mobile usage charge inquiry" as shown in FIG.

As described above, according to the embodiment of the present invention, it is possible to understand the questions of the natural language based customers through the machine learning-based system, and to use the question intention and the reliability value, which are the result values of the system, Can be provided.

In addition, it is possible to improve the processing and understanding of natural language processing and improve the development and operation efficiency by learning and automating the learning through the machine learning system without manually defining the corresponding pattern expression in AIML for the function of understanding the natural language.

In addition, realization of continuous conversation, context management, and scenario, which are the advantages of AIML, can be provided and a more complete chatbot service can be provided through interworking with a machine learning system using a large learning DB.

Although the machine learning server 300 is described as being a component in the chatbot system 10, the technical scope of the present invention is not limited thereto. For example, cloud service, which is a service for downloading various data after storing it in an external cloud server instead of an internal storage space such as a user's computer or a smart phone, is getting popular. Such a cloud service is a service providing artificial intelligence But also by an external cloud server or the like. For example, IBM's artificial intelligence technology, Watson and Google's TesnorFlow, are well known as open software. They use artificial intelligence to understand natural language questions, analyze and synthesize relevant data from vast amounts of data, and provide appropriate answers. They are providing cloud services that can use artificial intelligence technology through API (Application Programming Interface).

Next, a chatbot server according to another embodiment of the present invention will be described.

FIG. 9 is a schematic block diagram of a chatbot server 600 according to another embodiment of the present invention.

Referring to FIG. 9, a chatbot server 600 according to another embodiment of the present invention includes a communicator 610, an input device 620, a memory 630, a processor 640, and a bus 650.

The communicator 610 performs signal transmission with external components. Specifically, the communicator 610 communicates with the user terminal 100 via the network 400 and also communicates with the machine learning server 300. [

The input device 620 receives an input from the outside. These inputs may be data inputs or command inputs.

The memory 630 stores a plurality of intention patterns defined in AIML form and a template corresponding to a plurality of intent patterns in order to respond to a user's question.

Thus, the input device 620 may receive an input for adding or changing an intention pattern in the memory 630 and a template corresponding to the intention pattern.

The memory 630 is also configured to store a set of codes, which code is used to control the processor 640 to perform the following operations. This operation includes receiving an inquiry word from the user terminal 100 via the communicator 610, searching for a pattern matching the received query word, determining whether a pattern matching the query term is retrieved, (610), receiving an intention and a reliability value provided from the machine learning server (300) via the communicator (610), determining whether the received reliability value is equal to or greater than a preset threshold value Searching for a pattern matched through a plurality of intention patterns stored in the memory 630 with respect to a received intention when the reliability value is equal to or greater than a predetermined threshold value; The template corresponding to the searched pattern is extracted from the memory 630 and provided to the user terminal 100 through the communicator 610 in response to the query Includes an operation for providing a user terminal 100 to operate, if the credit value is less than a preset threshold response detection failure response through the communication device 610.

Optionally, the above operations further include adding or modifying an intent pattern and a template in memory 630 in accordance with information input via input device 620. [

The memory 630 may include a read only memory (ROM), a random access memory (RAM), and a non-volatile random access memory (NVRAM). The processor 640 may also be referred to as a controller, a microcontroller, a microprocessor, a microcomputer, or the like. In addition, the processor 540 may be implemented by hardware or firmware, software, or a combination thereof.

The bus 650 is configured to combine all the components of the chatbot server 600, i.e., the communicator 610, the input device 620, the memory 630, and the processor 640.

The embodiments of the present invention described above are not implemented only by the apparatus and method, but may be implemented through a program for realizing the function corresponding to the configuration of the embodiment of the present invention or a recording medium on which the program is recorded.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, It belongs to the scope of right.

Claims (15)

As a chatbot system based on AIML (Artificial Intelligence Markup Language)
A data set composed of a question and an intention corresponding to the question, a plurality of data sets are machine-learned to generate a learning model for detecting an intention corresponding to the question, and the intention corresponding to the question is detected A machine learning server; And
Storing a plurality of intention patterns defined in an AIML form and a template corresponding to the intention pattern by responding to a user's query and determining an intention corresponding to the query word through the machine learning server for a query term received from the user terminal Searching for an intent pattern corresponding to the intention that has been detected, and providing a template corresponding to the retrieved intent pattern to the user terminal in response to the query term,
. The method according to claim 1,
The chatbot server,
An intention-based database for storing the plurality of intention patterns and templates corresponding to the intention patterns;
A pattern matching unit for searching for a pattern matching the intention identified through the machine learning server; And
A template extracting unit for extracting a template corresponding to the pattern retrieved by the pattern matching unit,
. 3. The method of claim 2,
Wherein the pattern matching unit performs pattern matching on the query word through the intention-based database, and when the matched pattern is not found, transfers the query term to the machine learning server so that the intention corresponding to the query word And receiving an intention corresponding to the query term by receiving,
Chambot system. The method of claim 3,
Wherein the machine learning server assigns a confidence value to the intention of the machine learning to the data set and stores the reliability value in correspondence with the learned intention in the learning model,
And detecting an intention and a confidence value corresponding to the query term through the learning model when the query term is received from the pattern matching unit,
Chambot system. 5. The method of claim 4,
Wherein the pattern matching unit comprises:
Performing pattern matching using an intention received from the machine learning server when a reliability value received from the machine learning server is equal to or greater than a preset threshold value,
And providing a response detection failure to the user terminal in response to the query if the reliability value is less than the preset threshold value.
Chambot system. 6. The method according to any one of claims 2 to 5,
Wherein the plurality of intention patterns stored in the intention-based database are defined only as words or sentences without using wildcards,
Chambot system. A method in which an AIML-based chatbot system provides a chat service to a user terminal,
And an intention-based database for storing a plurality of intention patterns defined in AIML form by responding to a user's query and a template corresponding to the intention pattern,
Receiving a query from the user terminal;
Receiving a query from the machine learning server after transmitting the query to a machine learning server, the machine learning server comprising a machine learning module for learning a plurality of sets of data as a data set comprising a question and an intention corresponding to the question, Generating a learning model for detecting the intention, and using the generated learning model to detect intent corresponding to the question;
Retrieving a pattern matching the intention received from the machine learning server using the intention-based database; And
Extracting a template corresponding to the searched pattern from the intention-based data face and providing the template to the user terminal as a response corresponding to the query term
And providing the service to the service provider. 8. The method of claim 7,
Between the step of receiving the query and the step of receiving the intention,
Searching for a pattern matching the query terms using the intention-based database; And
Performing a step of providing, as a response corresponding to the query, if a pattern matching the query term is retrieved from the intention-based database, or
Performing a step of receiving the intention if a pattern matching the query term is not retrieved from the intention based database
Further comprising the steps of: 8. The method of claim 7,
Further comprising: receiving, from the machine learning server, a reliability value corresponding to the intention in the step of receiving the intention,
Before the step of searching for the pattern,
Determining whether the reliability value is greater than or equal to a preset threshold value; And
Performing a step of searching for the pattern when the reliability value is equal to or greater than the preset threshold value or providing a response detection failure to the user terminal in response to the query if the reliability value is smaller than the preset threshold value step
Further comprising the steps of: As a chatbot system based on AIML (Artificial Intelligence Markup Language)
An intention-based database for storing a plurality of intention patterns defined in AIML form by responding to a user's query and a template corresponding to the intention pattern;
A plurality of data sets are transmitted to an external cloud server as a data set composed of a question and an intention corresponding to the question to perform machine learning for intent detection corresponding to a question, A pattern matching unit for delivering the pattern to the cloud server to receive an intention corresponding to the query term and searching for a pattern matching the received intention, the external cloud server being a server capable of machine learning; And
A template extracting unit for extracting a template corresponding to the pattern retrieved by the pattern matching unit,
. 11. The method of claim 10,
Wherein the pattern matching unit performs pattern matching on the query word through the intention-based database, and when the matched pattern is not searched, the pattern matching unit transmits the query word to the external cloud server, Receiving intent,
Chambot system. 11. The method of claim 10,
Wherein the pattern matching unit further receives a reliability value for the intention from the external cloud server,
And performing pattern matching using an intention received from the external cloud server when the reliability value is equal to or greater than a preset threshold value.
Chambot system. As a chatbot system based on AIML (Artificial Intelligence Markup Language)
A communication device, a memory, and a processor,
Wherein the communicator comprises a user terminal and a machine learning server, wherein the machine learning server generates a learning model for machine learning a plurality of sets of data as a data set comprising a question and an intention corresponding to the question to detect an intention corresponding to the question And a server for detecting an intention corresponding to a question using the generated learning model,
Wherein the memory is configured to store a plurality of intention patterns defined in AIML form and a template corresponding to a plurality of intent patterns to respond to a user &apos; s query, and to store a set of codes,
The code includes:
Transmitting a query term received from the user terminal through the communicator to the machine learning server via the communicator;
Receiving an intention provided from the machine learning server via the communicator;
Retrieving from the memory an intent pattern that is matched against the intent;
Extracting a template corresponding to the retrieved intent pattern from the memory; And
And providing the extracted template to the user terminal through the communicator
Lt; RTI ID = 0.0 &gt; a &lt; / RTI &gt; processor,
Chambot system. 14. The method of claim 13,
The processor comprising:
Searching in the memory for a pattern matched for the query term; And
And transmitting the query term to the machine learning server via the communicator when a pattern matching the query word is not found
And further executing the chatbot system. 14. The method of claim 13,
The processor comprising:
Further receiving a reliability value corresponding to an intention received from the machine learning server;
Determining whether the reliability value is greater than or equal to a preset threshold value; And
Performing pattern matching on an intention received from the machine learning server when the reliability value is equal to or greater than the preset threshold value
And further executing the chatbot system.

Images