KR102468679B1 - Server, method and computer program for providing dialogue service using chatbot that remembers dialogue with user - Google Patents

Abstract

A server for providing a dialogue service using a chatbot which remembers a dialogue with a user comprises: a summary information generation unit for generating summary information in which a dialogue between a user terminal and the chatbot is summarized via a first trained model; a personal database storing the summary information for each user terminal; a sentence generation unit for obtaining, while the user terminal and the chatbot perform a real-time dialogue, relevant summary information related to the real-time dialogue from the personal database, and for generating a chatbot-side sentence by using the real-time dialogue and the relevant summary information via a second trained model; and a dialogue service providing unit for providing the real-time dialogue by using a user-side sentence received from the user terminal and the chatbot-side sentence. Accordingly, a dialogue between a user and a chatbot can be remembered for a long time to provide a consistent response when the user and the chatbot have a dialogue on a same subject. This application is a result of project support "Commercialization of Excellent Research from 2021 Artificial Intelligence Online Competition" within Open Competition Platform Construction Project hosted by the Ministry of Science and ICT and supported by the National IT Industry Promotion Agency using the Information and Communication Promotion Fund of the Ministry of Science and ICT [Project Number: R-20210726-011600].

Description

Server, method and computer program for providing a conversation service through a chatbot that remembers a conversation with a user

The present invention relates to a server, method, and computer program for providing a conversation service through a chatbot that remembers a conversation with a user.

A chatbot refers to a system implemented to respond to a user through a messenger based on predetermined response rules. Chatbots are based on artificial intelligence (AI) and big data analysis technology for smooth conversation, pattern recognition technology that allows machines to identify voices/texts, etc., and natural language that recognizes human language into computers and uses them for question answering and translation. It uses processing technology, semantic web technology that allows computers to understand information and reason logically, text mining technology that finds useful information from data composed of characters, and context-aware computing technology that understands the situation and context of the conversation partner.

Chatbots with these various technologies are mainly responsible for answering questions from consumers in home shopping, internet shopping malls, insurance companies, banks, food delivery, and accommodation reservations through messengers, and provide reliable, high-quality information. has

However, conventional chatbots do not have the ability to read and interpret sentences written by users and conduct conversations, so they often respond with incorrect sentences. In addition, chatbots have a disadvantage in that they cannot provide consistent responses to conversations with users because they do not remember conversations with users for a long time.

This application is sponsored by the Ministry of Science and ICT and supported by the National IT Industry Promotion Agency with the Information and Communication Promotion Fund of the Ministry of Science and ICT. This is the result obtained and performed [Task number: R-20210726-011600].

Korean Patent Publication No. 2018-0019869 (published on February 27, 2018)

It is intended to provide a conversation service providing server, method, and computer program that generate summary information summarizing conversations between user terminals and chatbots through a first learning model and store the summary information for each user terminal in an individual database.

While the user terminal and the chatbot are performing real-time conversation, obtain summary information related to the real-time conversation from the individual database, generate chatbot-side sentences using the real-time conversation and related summary information through the second learning model, and from the user terminal It is intended to provide a conversation service providing server, method, and computer program that provide real-time conversation using received user-side sentences and chatbot-side sentences.

However, the technical problem to be achieved by the present embodiment is not limited to the technical problems described above, and other technical problems may exist.

As a means for achieving the above-described technical problem, an embodiment of the present invention, a summary information generator for generating summary information summarizing a conversation between a user terminal and a chatbot through a first learning model, the summary for each user terminal A personal database for storing information, obtaining summary information related to the real-time conversation from the personal database while the user terminal and the chatbot are performing real-time conversation, and the real-time conversation and the related summary information through a second learning model It is possible to provide a conversation service providing server including a sentence generator for generating chatbot-side sentences using a sentence generator and a conversation service provider for providing the real-time dialogue using the user-side sentences received from the user terminal and the chatbot-side sentences. .

Another embodiment of the present invention, generating summary information summarizing the conversation between the user terminal and the chatbot through a first learning model, storing the summary information for each user terminal in an individual database, the user terminal and the While the chatbot is performing a real-time conversation, acquiring related summary information related to the real-time conversation from the individual database, and generating a chatbot-side sentence using the real-time conversation and the related summary information through a second learning model; and It is possible to provide a conversation service providing method including providing the real-time conversation by using the user-side sentence and the chatbot-side sentence received from the user terminal.

Another embodiment of the present invention, when the computer program is executed by a computing device, generates summary information summarizing the conversation between the user terminal and the chatbot through the first learning model, and stores the summary information for each user terminal in an individual database. Stores, while the user terminal and the chatbot are performing real-time conversation, obtaining related summary information related to the real-time conversation from the individual database, and using the real-time conversation and the related summary information through a second learning model It is possible to provide a computer program stored in a computer readable recording medium including a sequence of instructions for generating a chatbot-side sentence and providing the real-time conversation using the user-side sentence received from the user terminal and the chatbot-side sentence. .

The above-described means for solving the problems is only illustrative and should not be construed as limiting the present invention. In addition to the exemplary embodiments described above, there may be additional embodiments described in the drawings and detailed description.

According to any one of the above-described problem solving means of the present invention, by generating summary information summarizing the conversation between the user terminal and the chatbot through the first learning model, and storing the summary information in an individual database for each user terminal, the user and It is possible to provide a conversation service providing server, method, and computer program that can provide a consistent response when a conversation between chatbots is remembered for a long time and later a conversation between a user and a chatbot on the same topic is provided.

While the user terminal and the chatbot are performing real-time conversation, obtain summary information related to the real-time conversation from the individual database, generate chatbot-side sentences using the real-time conversation and related summary information through the second learning model, and from the user terminal It is possible to provide a conversation service providing server, method, and computer program for providing real-time conversation using the received user-side sentences and chatbot-side sentences.

1 is a configuration diagram of a conversation service providing server according to an embodiment of the present invention.
2 is an exemplary diagram for explaining a process of generating a plurality of conversation groups including a plurality of sentences included in a real-time conversation according to an embodiment of the present invention.
3A to 3C are exemplary diagrams for explaining a process of providing a conversation service through a chatbot that remembers a conversation with a user according to an embodiment of the present invention.
4 is a flowchart of a method of providing a conversation service through a chatbot that stores a conversation with a user performed in a conversation service providing server according to an embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail so that those skilled in the art can easily practice the present invention with reference to the accompanying drawings. However, the present invention may be embodied in many different forms and is not limited to the embodiments described herein. And in order to clearly explain the present invention in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification.

Throughout the specification, when a part is said to be "connected" to another part, this includes not only the case where it is "directly connected" but also the case where it is "electrically connected" with another element interposed therebetween. . In addition, when a part "includes" a certain component, this means that it may further include other components, not excluding other components, unless otherwise stated, and one or more other characteristics. However, it should be understood that it does not preclude the possibility of existence or addition of numbers, steps, operations, components, parts, or combinations thereof.

In this specification, a "unit" includes a unit realized by hardware, a unit realized by software, and a unit realized using both. Further, one unit may be realized using two or more hardware, and two or more units may be realized by one hardware.

In this specification, some of the operations or functions described as being performed by a terminal or device may be performed instead by a server connected to the terminal or device. Likewise, some of the operations or functions described as being performed by the server may also be performed in a terminal or device connected to the corresponding server.

Hereinafter, an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1 is a configuration diagram of a conversation service providing server according to an embodiment of the present invention. Referring to FIG. 1 , the conversation service providing server 100 includes a summary information generator 110, a filtering unit 120, an individual database 130, a summary information extractor 140, a sentence generator 150, a conversation It may include a service providing unit 160 and a learning unit 170 .

The summary information generation unit 110 may generate summary information summarizing the conversation between the user terminal and the chatbot through the first learning model.

To this end, the summary information generating unit 110 may create a plurality of conversation groups including a plurality of valid sentences. In this case, each of the plurality of conversation groups may include a plurality of chronologically continuous valid sentences among valid sentences sequentially stored in the buffer. Here, a plurality of valid sentences may be determined using pre-learned machine learning and deep learning models or rule-based models, but is not limited thereto.

The filtering unit 120 may determine whether sentences sequentially stored in the buffer are valid sentences in a real-time conversation, and may delete sentences other than valid sentences from the buffer. For example, when sentences such as "what did you eat for lunch?", "I ate tteokbokki", "hahahahaha", and "it was so delicious" are sequentially stored in the buffer, the filtering unit 120 filters "what did you eat for lunch?" Hahahaha" (eg, onomatopoeia, mimicry, etc.) can be deleted from the buffer.

Thereafter, while the real-time conversation is being performed, the summary information generating unit 110 may create a plurality of conversation groups including a plurality of sentences included in the real-time conversation, and generate summary information of each of the plurality of conversation groups. Here, the summary information is a summary that summarizes not only the user's personal information (eg, age, occupation, place of work, etc.) from real-time conversations, but also the user's hobbies, tendencies, habits, thoughts, and the user's thoughts on a specific topic. can The chatbot memorizes conversations related to the user based on the summary information, so that when a chatbot and the user later have a conversation on the same topic or the same keyword, the chatbot can communicate with the user like a human based on the summary information. can make it possible

Here, each sentence of the real-time conversation is sequentially stored in the buffer, and each of the plurality of conversation groups may include a plurality of temporally consecutive sentences among sentences sequentially stored in the buffer.

Also, the summary information generation unit 110 may update the summary information by grouping at least two or more conversation groups over time and further summarizing the at least two or more conversation groups.

For example, the updated summary information may include summary information consisting only of conversation topics or key keywords.

The summary information generating unit 110 may generate a plurality of conversation groups including a plurality of sentences through Equation 1 below, based on each sentence of the real-time conversation sequentially stored in the buffer.

Referring to Equation 1, n may represent the number of sentences in which conversation data for learning is input to the second learning model, and m may represent the number of sentences stored in the buffer. For example, when the conversation data for learning is designed such that 10 sentences are input to the second learning model, the summary information generating unit 110 groups each sentence of the real-time conversation sequentially stored in the buffer by 8 to make a plurality of sentences. You can create a chat group of Here, 0.8 may be changed according to experimental or simulation results, so it is not limited thereto.

The summary information generating unit 110 may generate a plurality of conversation groups so that two temporally adjacent conversation groups include a predetermined number of identical sentences (eg, three sentences). The summary information generator 110 may generate a plurality of chat groups so that two temporally adjacent chat groups include a predetermined number of identical sentences through Equation 2 below, for example.

Referring to Equation 2, n may represent the number of sentences in which conversation data for learning is input to the second learning model, and o may represent the number of identical sentences included in both temporally adjacent conversation groups. For example, when the conversation data for learning is designed to input, for example, 10 sentences to the second learning model, the summary information generating unit 110 determines that two conversation groups that are temporally adjacent to each other receive, for example, the same three sentences. can include Here, 0.3 may be changed according to experimental or simulation results, so it is not limited thereto.

For example, the summary information generating unit 110 sets turns of a plurality of sentences included in the first conversation group so that three identical sentences are included in both the first and second conversation groups that are temporally adjacent, for example, [0, 1, 2, 3, 4, 5, 6, 7], and the turns of a plurality of sentences included in the second dialogue group are, for example, [5, 6, 7, 8, 9, 10, 11, 12] to create a plurality of chat groups.

Hereinafter, a process of generating a plurality of conversation groups including a plurality of sentences included in a real-time conversation while a real-time conversation is being performed will be described with reference to FIG. 2 .

2 is an exemplary diagram for explaining a process of generating a plurality of conversation groups including a plurality of sentences included in a real-time conversation according to an embodiment of the present invention.

Referring to FIG. 2 , while the real-time conversation 200 is being performed, the summary information generating unit 110 creates a plurality of conversation groups including a plurality of sentences included in the real-time conversation 200, and generates a plurality of conversation groups for each of the plurality of conversation groups. Summary information can be created.

To this end, the filtering unit 120 may sequentially delete sentences other than valid sentences from the real-time conversation 200 in the buffer. For example, the filtering unit 120 may delete meaningless sentences such as 'ㅠㅠㅠ' 211, which are not valid sentences including meaningful contents or meaningful keywords in the real-time conversation 200.

Thereafter, the summary information generator 110 may create a plurality of conversation groups to include a plurality of chronologically continuous valid sentences among valid sentences sequentially stored in the buffer. Here, the summary information generating unit 110 may create a plurality of conversation groups by grouping eight sentences of the real-time conversation 200 sequentially stored in the buffer. For example, the summary information generating unit 110 may create the first conversation group 210 from "what did you eat for lunch?" of the chatbot to "make sure to bring an umbrella" of the chatbot.

Also, the summary information generator 110 may further generate a second conversation group 220 including a plurality of valid sentences temporally continuous with the first conversation group 210 from the real-time conversation 200 . Here, the summary information generating unit 110 may generate a plurality of conversation groups such that two temporally adjacent conversation groups overlap with a predetermined number of sentences (eg, three sentences).

For example, when the first conversation group 210 includes sentences from "What did you eat for lunch?" of the chatbot to "Be sure to take an umbrella" of the chatbot, the summary information generator 110 generates the first conversation group ( 210), the user's sentence "I'm going to go for a walk in the park today, because the weather is nice", and the chatbot's "Make sure to bring an umbrella" sentence are also included in the second conversation group 220, and the user's "Yes. Later. Sentences up to "I'll go for a walk and drink a cup of warm citron tea" can be generated as the second conversation group 220 .

Through this process, by creating a plurality of conversation groups, it is possible to have conversation connectivity between each conversation group.

Returning to FIG. 1 , the individual database 130 may store summary information for each user terminal.

The summary information extractor 140 may extract related summary information from among a plurality of pieces of summary information stored in a database corresponding to the user terminal while a real-time conversation is performed through the third learning model. Here, the third learning model is BERT (Bidirectional Encoder Representations form Transformer), which is a model that trains the model in advance with a large amount of unlabeled data and then performs transfer learning with labeled data having a specific task. An encoder model such as RoBERTa (Robustly Optimized BERT Pretraining Approach) in which hyperparameters and training data size are adjusted may be used, but is not limited thereto. Also, the third learning model may be a model fine-tuned to a pretrained language model (PLM) as a summary information retrieval task.

The summary information extractor 140 selects at least one piece of related summary information (eg, m pieces) from among a plurality of pieces of summary information stored from a database corresponding to the user terminal while a real-time conversation is performed through the third learning model. can be extracted. Here, as the third learning model, a plurality of Ranker models may be used to select a plurality of related summary information among a plurality of summary information, but is not limited thereto.

For example, after weighting the latest data among a plurality of pieces of summary information, deriving a ranking based on scores of each of the pieces of weighted pieces of summary information, and then extracting a plurality of related pieces of summary information equal to or higher than a preset ranking rank. can

For another example, a core keyword may be derived from among a plurality of pieces of summary information, and a plurality of pieces of related summary information including a keyword similar to the core keyword may be extracted.

The sentence generation unit 150 obtains summary information related to real-time conversation from an individual database while the user terminal and the chatbot are performing real-time conversation, and uses the real-time conversation and related summary information through a second learning model to generate sentences on the chatbot side. can create

The sentence generating unit 150 may generate a chatbot-side sentence by inputting a temporally latest conversation group and related summary information among a plurality of sentences included in a real-time conversation to the second learning model. Here, the second learning model is a model capable of exchanging basic conversations between the user terminal and the chatbot, and can generate sentences related to summary information. Any model may be used as the second learning model as long as it is a learning model that uses real-time conversation and summary information as inputs. In addition, when a user-side sentence is received from the user terminal, the second learning model may generate a chatbot-side next sentence related to the user-side sentence.

The conversation service provider 160 may provide real-time conversation using the user-side sentence and the chatbot-side sentence received from the user terminal.

The learning unit 170 may learn the first learning model to generate summary information based on effective sentences from the conversation data for learning by inputting the conversation data for learning to the first learning model.

The learning unit 170 may learn the second learning model to generate chatbot-side sentences related to the learning conversation data by inputting the learning conversation data and learning-related summary information related to the learning conversation data to the second learning model. Here, the learning unit 170 may input learning dialogue data (eg, 10 sentences each) and learning summary information related to the learning dialogue data to the second learning model as a data set.

The learning unit 170 may learn the third learning model to input the conversation data for learning and the summary information for learning to the third learning model to extract related summary information related to the conversation data for learning.

3A to 3C are exemplary diagrams for explaining a process of providing a conversation service through a chatbot that remembers a conversation with a user according to an embodiment of the present invention.

Referring to FIG. 3A , it is assumed that the chatbot 300 has a real-time conversation with a user, Samantha 310. For example, to the chatbot 300's question "How old are you?", Samantha 310 responds "I'm 23 years old", and to the chatbot 300's question "What are you doing?", Samantha ( 310) responds "I am eating pizza", and to the chatbot 300's question "Are you a tourist?", Samantha 310 responds "No, I am studying at the library", and the chatbot 300 ) to the question "Where do you work?", suppose Samantha (310) responds "I work at a Seoul hospital".

Several hours later, when Samantha 310 asks the chatbot 300, "Where do I work?", the chatbot 300 extracts 'occupation and location' from a plurality of stored summary information related to previous conversations and "Seoul Hospital" may be answered based on relevant summary information. In addition, when Samantha 310 asks, “Do you remember how old I am?”, the chatbot 300 responds with “Age” related summary information extracted from a plurality of pieces of summary information stored in relation to previous conversations. 23 years old." In addition, when Samantha 310 asks, “What am I doing?”, the chatbot 300 responds with “current and working” related summary information extracted from a plurality of stored summary information related to previous conversations. I'm studying at the library."

Referring to FIG. 3B , assume that a chatbot 320 has a real-time conversation with a current user 330 . For example, to the chatbot 320's question "Which hamburger do you like best?", the present 330 responds "My favorite hamburger is a mushroom wafer haha", and the chatbot 320 responds "Me too" Suppose you responded with something like "I like it... it looks delicious...".

After 3 days, when the chatbot 320 has a real-time conversation with the current user 330 again, the chatbot 320 responds with "hamburger!" to the question "What did you eat for lunch?" In one case, the chatbot 320 is based on summary information related to 'your favorite hamburger-mushroom wafer' extracted from a plurality of pieces of summary information stored in relation to previous conversations, "Is it a mushroom wafer? Is that your favorite?" Respond together, and the present 330 may respond with "That's right haha that's the most delicious".

Referring to FIG. 3C , it is assumed that the chatbot 340 has a real-time conversation with a user Ryan 350. For example, to the chatbot 340's question "What is your job?", Ryan 350 responds "I'm a software engineer", and to the chatbot's 340's question "What program did you develop?" , Assume that Ryan 350 responds with "java, python, c#, c++".

Thereafter, when the chatbot 340 responds to the question, "But what did you say your job was?", when Ryan 350 jokingly replies to the chatbot 340, "I'm a teacher," the chatbot 340 responds to the previous conversation and By extracting the relevant summary information related to 'occupation-soft engineer' among the related summary information, a response such as "You said you were an engineer" can be made.

Conventionally, a conversation between a user and a chatbot is not stored for a long time, but only while the conversation is initiated. In addition, since the conventional chatbot does not have the ability to memorize conversation contents, it is not possible to provide consistent responses, and there are cases in which the contents shared in the previous conversation are re-queried. As a result, users did not feel as if they were having a conversation between people.

However, according to the present invention, the conversation between the user and the chatbot is stored as summary information in an individual database, rather than being remembered only while the conversation between the user and the chatbot is initiated, so that the conversation between the user and the chatbot can be remembered for a long time. In addition, by extracting summary information related to a specific keyword or a specific topic from among a plurality of summary information stored in previous conversations between the chatbot and the user, consistent responses can be provided to the user over time. In addition, the chatbot may remember the user's taste and tendency and provide a customized response thereto.

It may be executed by a computer program stored in a medium including a sequence of instructions for providing a conversation service through a chatbot that stores a conversation with a user in the conversation service providing server. When the computer program is executed by a computing device, the first learning model generates summary information summarizing the conversation between the user terminal and the chatbot, stores the summary information for each user terminal in an individual database, and allows the user terminal and the chatbot to communicate in real time While performing, obtain related summary information related to real-time conversation from the individual database, generate chatbot-side sentences using real-time conversation and related summary information through the second learning model, and user-side sentences received from the user terminal and chatbot-side It can include a sequence of instructions that allow you to provide real-time conversations using sentences.

4 is a flowchart of a method of providing a conversation service through a chatbot that stores a conversation with a user performed in a conversation service providing server according to an embodiment of the present invention. Referring to FIG. 4, a method of providing a conversation service through a chatbot that remembers a conversation with a user performed in a conversation service providing server 100 is processed time-sequentially according to the embodiments shown in FIGS. 1 to 3C. contains the steps Therefore, even if the content is omitted below, it is also applied to the method of providing a conversation service through a chatbot that remembers a conversation with a user performed in the conversation service providing server 100 according to the embodiment shown in FIGS. 1 to 3C.

In step S410, the conversation service providing server 100 may generate summary information summarizing the conversation between the user terminal and the chatbot through the first learning model.

In step S420, the conversation service providing server 100 may store summary information for each user terminal in the individual database.

In step S430, the conversation service providing server 100 obtains summary information related to real-time conversation from an individual database while the user terminal and the chatbot are performing real-time conversation, and uses the real-time conversation and related summary information through a second learning model. In this way, the chatbot-side sentence can be generated.

In step S440, the conversation service providing server 100 may provide real-time conversation using the user-side sentence and the chatbot-side sentence received from the user terminal.

In the foregoing description, steps S410 to S440 may be further divided into additional steps or combined into fewer steps, depending on the implementation of the present invention. Also, some steps may be omitted as needed, and the order of steps may be switched.

The method for providing a conversation service through a chatbot that remembers a conversation with a user performed in a conversation service providing server described with reference to FIGS. 1 to 4 is a computer program stored in a medium executed by a computer or instructions executable by a computer It may also be implemented in the form of a recording medium including. In addition, the method of providing a conversation service through a chatbot that remembers a conversation with a user performed in a conversation service providing server described with reference to FIGS. 1 to 4 is implemented in the form of a computer program stored in a medium executed by a computer It can be.

Computer readable media can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media. Also, computer readable media may include computer storage media. Computer storage media includes both volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data.

The above description of the present invention is for illustrative purposes, and those skilled in the art can understand that it can be easily modified into other specific forms without changing the technical spirit or essential features of the present invention. will be. Therefore, the embodiments described above should be understood as illustrative in all respects and not limiting. For example, each component described as a single type may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined form.

The scope of the present invention is indicated by the following claims rather than the detailed description above, and all changes or modifications derived from the meaning and scope of the claims and equivalent concepts should be construed as being included in the scope of the present invention. do.

100: conversation service providing server
110: summary information generating unit
120: filtering unit
130: individual database
140: summary information extraction unit
150: sentence generator
160: conversation service provider
170: learning unit

Claims (20)

In a server that provides a conversation service through a chatbot that remembers a conversation with a user,
Summary information generation unit for generating summary information summarizing the conversation between the user terminal and the chatbot through the first learning model;
an individual database for storing the summary information for each user terminal;
While the user terminal and the chatbot are performing a real-time conversation, obtaining related summary information related to the real-time conversation from the individual database, and using the real-time conversation and the related summary information through a second learning model to construct a sentence on the chatbot side a sentence generation unit to generate; and
A conversation service provider for providing the real-time conversation using the user-side sentence and the chatbot-side sentence received from the user terminal
Including,
The summary information generator generates a plurality of conversation groups including a plurality of sentences included in the real-time conversation while the real-time conversation is being performed, and generates the summary information of each of the plurality of conversation groups;
Each sentence of the real-time conversation is sequentially stored in a buffer;
Each of the plurality of conversation groups includes the plurality of sentences sequentially sequentially stored in the buffer, and
A filtering unit for determining whether sentences sequentially stored in the buffer are valid sentences in the real-time conversation, and deleting sentences other than the valid sentences from the buffer.
Further comprising, a conversation service providing server.
According to claim 1,
A summary information extraction unit extracting the related summary information from among a plurality of summary information stored in a database corresponding to the user terminal while the real-time conversation is being performed through a third learning model.
Further comprising, a conversation service providing server.
delete delete According to claim 1,
The conversation service providing server of claim 1 , wherein the summary information generation unit generates the plurality of conversation groups so that two conversation groups that are temporally adjacent to each other include a predetermined number of identical sentences.
delete According to claim 1,
The summary information generation unit generates a plurality of conversation groups including a plurality of the valid sentences;
Wherein each of the plurality of conversation groups includes a plurality of chronologically continuous valid sentences among valid sentences sequentially stored in the buffer.
According to claim 1,
Wherein the sentence generation unit generates the chatbot-side sentence by inputting the temporally latest conversation group and the related summary information among the plurality of sentences included in the real-time conversation to the second learning model, a conversation service providing server .
According to claim 2,
Further comprising a learning unit for learning the second learning model to input learning conversation data and learning-related summary information related to the learning conversation data to the second learning model to generate a chatbot-side sentence related to the learning conversation data. Serving server.
A method for providing a conversation service through a chatbot that remembers a conversation with a user in a conversation service providing server,
Generating summary information summarizing a conversation between a user terminal and a chatbot through a first learning model;
storing the summary information for each user terminal in an individual database;
While the user terminal and the chatbot are performing a real-time conversation, obtaining related summary information related to the real-time conversation from the individual database, and using the real-time conversation and the related summary information through a second learning model to construct a sentence on the chatbot side generating; and
Providing the real-time conversation using the user-side sentence and the chatbot-side sentence received from the user terminal
Including,
Generating the summary information,
generating a plurality of conversation groups including a plurality of sentences included in the real-time conversation while the real-time conversation is being performed; and
generating the summary information of each of the plurality of conversation groups;
Each sentence of the real-time conversation is sequentially stored in a buffer;
Each of the plurality of conversation groups includes the plurality of sentences sequentially sequentially stored in the buffer, and
The method of providing a conversation service further comprising determining whether sentences sequentially stored in the buffer are valid sentences in the real-time conversation.
According to claim 10,
Extracting the related summary information from among a plurality of pieces of summary information stored in a database corresponding to the user terminal while the real-time conversation is being performed through a third learning model.
Further comprising, a conversation service providing method.
delete delete According to claim 10,
Generating the summary information,
and generating the plurality of conversation groups such that both of the two conversation groups that are temporally adjacent contain a predetermined number of identical sentences.
delete According to claim 10,
The method further comprising deleting sentences other than the valid sentences from the buffer.
According to claim 10,
Generating the summary information,
generating a plurality of the conversation groups including a plurality of the valid sentences; and
The method of claim 1 , wherein each of the plurality of conversation groups includes a plurality of chronologically continuous valid sentences among valid sentences sequentially stored in the buffer.
According to claim 10,
The step of generating the chatbot-side sentence,
and generating the chatbot-side sentence by inputting the conversation group and related summary information, which are temporally latest among the plurality of sentences included in the real-time conversation, into the second learning model. .
According to claim 11,
Learning the second learning model to generate chatbot-side sentences related to the learning conversation data by inputting learning conversation data and learning-related summary information related to the learning conversation data to the second learning model; How We Provide Our Services.
A computer program stored in a computer-readable recording medium comprising a sequence of instructions for providing a conversation service through a chatbot that remembers a conversation with a user,
When the computer program is executed by a computing device,
Generating summary information summarizing the conversation between the user terminal and the chatbot through the first learning model;
Storing the summary information for each user terminal in an individual database;
While the user terminal and the chatbot are performing a real-time conversation, obtaining related summary information related to the real-time conversation from the individual database, and using the real-time conversation and the related summary information through a second learning model to construct a sentence on the chatbot side create,
Providing the real-time conversation using the user-side sentence and the chatbot-side sentence received from the user terminal;
While the real-time conversation is being performed, a plurality of conversation groups including a plurality of sentences included in the real-time conversation are created, and the summary information of each of the plurality of conversation groups is created;
Each sentence of the real-time conversation is sequentially stored in a buffer;
Each of the plurality of conversation groups includes the plurality of sentences sequentially sequentially stored in the buffer, and
A computer program stored in a computer-readable recording medium comprising a sequence of instructions for determining whether sentences sequentially stored in the buffer are valid sentences in the real-time conversation.

Images