KR102098003B1 - Method, apparatus and computer-readable medium for operating chatbot - Google Patents

Abstract

According to an embodiment of the present invention, disclosed is a chatbot driving method. Operations stored in a computer program to implement the method comprise: determining a reference sentence having a similarity to an input sentence that is greater than or equal to a threshold value wherein the similarity is received from a user terminal; determining a sub-topic of the reference sentence and determining a conversation topic based on the sub-topic of the reference sentence; determining an answer based on a first sub-topic corresponding to a next sub-topic of the sub-topic of the reference sentence among a plurality of sub-topics included in the conversation topic; and transmitting the answer to the user terminal.

Description

METHOD, APPARATUS AND COMPUTER-READABLE MEDIUM FOR OPERATING CHATBOT

The present disclosure relates to a technology for driving a chatbot, and more particularly, to a method in which the chatbot can lead a conversation.

In recent years, as the society has become complicated and the anxiety about the future has increased, the number of people insured has increased significantly. Various financial products have been developed to meet such demands, and financial products targeting leisure activities or leisure activities of insurance subscribers have been proposed as leisure time such as five-day work week increases.

In this regard, various types of insurance products are provided by various insurance companies, and the insurer can select an insurance company and products that offer favorable conditions to the contract. In addition, the insurance company may recommend products suitable for the insurance subscriber.

Korean Patent Publication No. KR20000030728A discloses a method of recommending and providing insurance that can incentivize a user to insure insurance products in accordance with positive, negative, negative, and five elements.

Meanwhile, insurance companies receive customer consultations and inquiries through insurance designers and counselors, and respond accordingly. For example, in the case of a potential customer who is interested in auto insurance, the system consists of a system that recommends insurance suitable for potential customers through consultation after contacting an insurance company counselor or insurance planner for insurance.

However, insurance inquiries through a designer or a counselor have a problem that is difficult to be processed immediately at a time desired by a customer because of manpower issues and limitations in consultation time. In addition, insurance architects and counselors who specialize in this also had the inconvenience of unnecessarily increasing work because they had to deal with even simple tasks one by one.

In addition, chatbots have a limitation in that they cannot receive the feeling of being consulted by an actual counselor, in a way that only a customer's question is answered.

Therefore, there may be a demand in the art for a method in which the chatbot can lead the conversation in a way that the customer is curious about through the chatbot.

The present disclosure has been devised in response to the above-mentioned background technology, and is intended to provide a method for a chatbot to lead a conversation in consultation with a user.

Disclosed is a computer program stored in a computer-readable storage medium in accordance with one embodiment of the present disclosure for realizing the above-described problems. When the computer program is executed on one or more processors of the computing device, the following operations are performed for driving the chatbot, and the operations are: determining a reference sentence having a similarity with an input sentence received from a user terminal having a threshold value or higher. ; Determining a sub-subject of the reference sentence, and determining a topic of a conversation based on the sub-subject of the reference sentence; Determining an answer based on a first sub-subject corresponding to a next sub-subject of the sub-subject of the reference sentence among a plurality of sub-subjects included in the topic of the conversation; And transmitting the answer to the user terminal.

In an alternative embodiment of computer program operations to perform the following operations for driving the chatbot, the answer may include a response for deriving an input sentence desired to be received from the user terminal.

In an alternative embodiment of computer program operations to perform the following operations for driving the chatbot, the operation of determining a reference sentence having a similarity level or higher with an input sentence received from a user terminal is greater than or equal to a threshold value. Obtaining an input sentence vector by calculating through a pre-processing model including a network function; Obtaining one or more reference sentence vectors by calculating each of the one or more reference sentences through the pre-processing model; And determining a reference sentence having a similarity level equal to or greater than a threshold value for each of the input sentence vector and the one or more reference sentence vectors.

In an alternative embodiment of computer program operations to perform the following operations for driving the chatbot, the operation of determining a reference sentence having a similarity level or higher with an input sentence received from a user terminal is greater than or equal to a threshold value. Obtaining an input sentence vector by calculating through a pre-processing model including a network function; Obtaining a past input sentence vector by calculating a past input sentence input before the input sentence through the pre-processing model; Determining a cumulative input vector based on the input sentence vector and the past input sentence vector; And determining the reference sentence having a degree of similarity with the accumulated input vector equal to or greater than a threshold value.

In an alternative embodiment of computer program operations to perform the following operations for driving the chatbot, the operation of determining the sub-subject of the reference sentence and determining the topic of the conversation based on the sub-subject of the reference sentence, Determining a reference past input sentence having a similarity level with a threshold value equal to or higher than a past input sentence input before the input sentence; Determining a sub-subject of the past input sentence; And determining a topic of conversation based on the sub-subject of the reference sentence and the sub-subject of the past input sentence.

In an alternative embodiment of computer program operations to perform the following operations for driving the chatbot, the operation of determining the sub-subject of the reference sentence and determining the topic of the conversation based on the sub-subject of the reference sentence, And determining a topic of the conversation based on the sub-subject of the reference sentence and the user's status information.

In an alternative embodiment of computer program operations to perform the following operations for driving the chatbot, the operation of determining a reference sentence having a similarity level above the threshold value for each of the input sentence vector and the one or more reference sentence vectors includes: Outputting one or more scoring values indicating the degree of similarity to each of the input sentence vector and the one or more reference sentence vectors; Determining a reference sentence based on a class of scoring values equal to or greater than a threshold value among the one or more scoring values; And when there is no class corresponding to a scoring value of a threshold value or more among the one or more scoring values, determining that the reference sentence cannot be determined.

In an alternative embodiment of computer program operations to perform the following operations for driving the chatbot, if there is no class corresponding to a scoring value above a threshold among the one or more scoring values, determining that the reference sentence cannot be determined The operation may include transmitting at least one of an answer to determine whether a reference sentence based on a class of the maximum scoring value among the one or more scoring values is correct and an answer that cannot be determined for the input sentence to the user terminal. .

In an alternative embodiment of computer program operations to perform the following operations for driving the chatbot, the operation of transmitting the answer to the user terminal includes: when the answer includes content, the input sentence and the answer The method may include transmitting an interface for a first layer including the chat window and a second layer including the content to the user terminal to display the chat window and the content together on the user terminal.

In an alternative embodiment of computer program operations to perform the following operations for driving a chatbot, a first sub corresponding to a next sub topic of a sub topic of the reference sentence among a plurality of sub topics included in the topic of the conversation The operation of determining an answer based on a subject may include comparing the first sub-subject and the sub-subject of the past input sentence input before the input sentence from the user terminal; And when the first sub-subject and the sub-subject of the past input sentence match, determining an answer based on the second sub-subject corresponding to the next sub-subject of the first sub-subject.

In an alternative embodiment of computer program operations to perform the following operations for driving the chatbot, if there are two or more topics obtained based on the sub-themes, receive a selection input for one of the two or more topics The method may further include transmitting an answer for the user terminal.

In an alternative embodiment of computer program operations to perform the following operations for driving the chatbot, a method of driving the chatbot according to another embodiment of the present disclosure is disclosed. The method includes the steps of determining a reference sentence having a similarity level or higher with an input sentence received from a user terminal; Determining a sub-subject of the reference sentence, and determining a topic of conversation based on the sub-subject of the reference sentence; And determining a response based on a first sub-subject corresponding to a next sub-subject of the sub-subject of the reference sentence among a plurality of sub-subjects included in the topic of the conversation. And transmitting the answer to the user terminal.

A computing device is disclosed in accordance with another embodiment of the present disclosure. The computing device includes a processor including one or more cores; And memory; Including, the processor, determines a reference sentence having a similarity or higher than a threshold value of the input sentence received from the user terminal, determines the sub-subject of the reference sentence, and the subject of the conversation based on the sub-subject of the reference sentence And determines an answer based on a first sub-subject corresponding to a next sub-subject of the sub-subject of the reference sentence among a plurality of sub-subjects included in the topic of the conversation, and transmits the response to the user terminal You can.

The present disclosure can provide a method of driving a chatbot.

1 is a block diagram of a computing device performing an operation for driving a chatbot according to an embodiment of the present disclosure.
2 is a diagram illustrating a flowchart between a server and a user terminal according to an embodiment of the present disclosure.
3 is a diagram illustrating subject and reference sentences stored in a server according to an embodiment of the present disclosure.
4 is a diagram illustrating an operation interface of a chatbot according to an embodiment of the present disclosure.
5 is a flowchart of a method for driving a chatbot according to an embodiment of the present disclosure.
6 is a block diagram illustrating logic for implementing a method of driving a chatbot according to an embodiment of the present disclosure.
7 is a block diagram of a computing device according to an embodiment of the present disclosure.

Various embodiments are now described with reference to the drawings, and like reference numbers throughout the drawings are used to indicate similar elements. In this specification, various descriptions are presented to provide an understanding of the present disclosure. However, it is clear that these embodiments can be practiced without these specific details. In other instances, well-known structures and devices are provided in block diagram form in order to facilitate describing the embodiments.

As used herein, the terms “component”, “module”, “system” and the like refer to computer-related entities, hardware, firmware, software, combinations of software and hardware, or execution of software. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, a thread of execution, a program, and / or a computer. For example, both the application and the computing device running on the computing device can be components. One or more components can reside within a processor and / or thread of execution, and one component can be localized within one computer, or distributed between two or more computers. In addition, these components can execute from various computer readable media having various data structures stored therein. The components are, for example, signals having one or more data packets (e.g., data from one component interacting with another component in a local system, distributed system, and / or data via another system and a network such as the Internet via a signal) ) Through local and / or remote processes.

Descriptions of the presented embodiments are provided to enable any person skilled in the art to use or practice the present disclosure. Various modifications to these embodiments will be apparent to those skilled in the art of the present disclosure, and the general principles defined herein can be applied to other embodiments without departing from the scope of the present disclosure. Thus, the present disclosure should not be limited to the embodiments presented herein, but should be interpreted in the broadest scope consistent with the principles and novel features presented herein.

In one embodiment of the present disclosure, the server may include other configurations for performing the server environment of the server. The server can include any type of device. The server is a digital device, and may be a digital device equipped with a processor and having computing power, such as a laptop computer, a notebook computer, a desktop computer, a web pad, and a mobile phone. The server may be a web server that processes services. The above-mentioned types of servers are only examples, and the present disclosure is not limited thereto.

Throughout this specification, models, neural networks, network functions, and neural networks may be used in the same sense. A neural network may consist of a set of interconnected computational units, which may generally be referred to as "nodes." These “nodes” may also be referred to as “neurons”. The neural network comprises at least one node. The nodes (or neurons) that make up the neural networks can be interconnected by one or more “links”.

Throughout this specification, a model can be used to mean including one or more network functions. The model can perform data processing along with other models. The models can be interconnected in series or in parallel with other models.

1 is a block diagram of a computing device performing an operation for driving a chatbot according to an embodiment of the present disclosure.

The computing device 100 for a method of driving a chatbot according to an embodiment of the present disclosure may be used to classify a customer's intention in a one-to-one conversation with a user about one of consultation, inquiry, or customer service related to a personal financial product. You can. Here, personal financial products may refer to general products such as insurance, loans, pensions, and funds.

The computing device 100 for a method of driving a chatbot according to an embodiment of the present disclosure may include a network unit 110, a processor 120, and a memory 130.

The network unit 110 may receive an input sentence including one or more words from the user terminal 200. Here, the input sentence may include a sentence received by the computing device 100 from the user terminal 200. The input sentence may be the first sentence that the computing device 100 receives from the user terminal 200 or may be a sentence after the first sentence. Further, the input sentence may be one or more sentences received until the computing device 100 performs an arbitrary operation on the user terminal 200. For example, the network unit 110 may receive the “Hello Chatbot” as the first input sentence received through the user terminal 200, and then receive the second input sentence “I want to have insurance consultation”. have. Here, when the computing device 100 does not perform the response to the first input sentence and the second input sentence or outputting a result value, the first input sentence and the second input sentence may be determined as one input sentence. The detailed description of the aforementioned input sentence is only an example, and the present disclosure is not limited thereto.

The network unit 110 may receive a past input sentence including one or more words from the user terminal 200. Here, the past input sentence may mean a sentence received before the computing device 100 receives the input sentence from the user terminal 200. The computing device 100 may determine similarity with reference sentences for the input sentence based on the past input sentence.

The network unit 110 may transmit a response response to the user terminal 200. Response The response may be a response to an input sentence received from the user terminal.

The network unit 110 may transmit an answer to the user terminal 200. The answer may be a response for deriving an input sentence desired to be received from the user terminal. The answer may include at least one of an answer to determine whether the reference sentence based on the score value is correct and an answer that determination of the input sentence is impossible. The answer may include an answer for receiving a selection input for one of two or more subjects identified based on a user's input sentence. The answer may include an answer based on a reference sentence of a sub-subject with the highest priority among one or more sub-subjects included in one subject identified based on a user's input sentence. More specifically, the network unit 110 may obtain an answer generated by the processor 120. Additionally, the network unit 110 may obtain an answer sentence from at least one of the financial company computing device groups. In addition, the network unit 110 may transmit the response to the user terminal 200 together with the interface through a chatbot connected to the user terminal 200. In the above process, the network unit 110 may communicate with the user terminal 200 through a wired / wireless network.

Additionally, the network unit 110 may transmit an answer matching the input sentence stored in the memory 130 to the user terminal 200.

The user terminal 200 may include a personal computer (PC), a notebook (note book), a mobile terminal (smart phone), a smart phone (smart phone), a tablet PC (tablet pc), VR (Virtual Reality) device, etc. It can include all types of terminals that can access the wired / wireless network. In addition, the network unit 110 may include a wired / wireless Internet module for network access. Wireless Internet technology (Wireless LAN) (Wi-Fi), Wibro (Wireless broadband), Wimax (World Interoperability for Microwave Access), HSDPA (High Speed Downlink Packet Access) may be used. As a wired Internet technology, digital subscriber line (XDSL), fibers to the home (FTTH), and power line communication (PLC) may be used.

In addition, the network unit 110 includes a short-range communication module, and is located at a relatively short distance and transmits and receives data to and from an electronic device (eg, an external terminal) including the short-range communication module. Bluetooth, radio frequency identification (RFID), infrared data association (IrDA), ultra wideband (UWB), ZigBee, etc. may be used as short range communication technology.

The processor 120 may include one or more cores, a central processing unit (CPU) of a computing device, a general purpose graphics processing unit (GPGPU), and a tensor processing unit (TPU). unit), and a processor for driving a chatbot. The processor 120 may read a computer program stored in the memory 130 and perform a method of driving a chatbot according to an embodiment of the present disclosure. According to an embodiment of the present disclosure, the processor 120 may perform calculation for a method of driving a chatbot.

At least one of the CPU, GPGPU, and TPU of the processor 120 may process an operation for an input sentence for a method of driving a chatbot, a similarity operation for an input sentence, and a reference sentence. For example, for the method of driving the chatbot, the CPU and the GPGPU may process an operation of determining a reference sentence having a similarity threshold or higher based on the input sentence vector and the past input sentence vector. In addition, in an embodiment of the present disclosure, similarity determination based on a scoring value for a method of driving a chatbot may be processed by using processors of a plurality of computing devices together. Further, the computer program executed in the computing device according to an embodiment of the present disclosure may be a CPU, GPGPU, or TPU executable program.

According to an embodiment of the present disclosure, a reference sentence may be determined based on an input sentence received from the user terminal 200.

The processor 120 may determine a reference sentence in which the similarity with the input sentence received from the user terminal 200 is greater than or equal to a threshold value.

More specifically, the processor 120 may obtain an input sentence vector by calculating the input sentence through the pre-processing model 300 including one or more network functions.

The pre-processing model 300 according to an embodiment of the present disclosure may perform natural language processing using a Recurrent Neural Network (RNN). Here, RNN may mean a deep learning model for learning a pattern based on data of a sentence including one or more words. That is, the processor 120 may repeatedly learn an input sentence based on an RNN as input data and an input sentence vector for the input sentence as output data. Accordingly, the processor 120 may perform more accurate vectorization of input sentences as input sentences received through the network unit 110 are accumulated and input sentence vector information for the input sentences is accumulated.

In addition, the RNN according to an embodiment of the present disclosure may solve a problem of vanishing gradient due to continuous learning through at least one of a Long Short-Term Memory Units (LSTM) method and a Gated Recurrent Units (GRU) method. . More specifically, the processor 120 may prevent gradient loss by using an LSTM method that adds input, output, and forget gates having respective weights to the basic RNN structure. In addition, in another embodiment, the processor 120 may prevent gradient loss in a GRU method (eg, dynamic Gated Recurrent Units (GRU) method) that additionally includes only input and forget gates in the RNN structure. The detailed description of the structure of the RNN described above is merely an example, and the present disclosure is not limited thereto.

More specifically, the processor 120 may convert the input sentence into an input sentence vector form by calculating through the pre-processing model 300 including one or more network functions. Here, the pre-processing model 300 may include a function for converting an input sentence into a set of one or more morphemes through morpheme analysis. In addition, the pre-processing model 300 may include a function for converting each of one or more morphemes into a vector of a predetermined number of dimensions (eg, 100 dimensions). For example, the pre-processing model may include a "Word2Vec" function. In addition, the pre-processing model 300 may calculate in the form of a weighting matrix using a function for calculating a distance between vectors converted in morpheme units. In addition, the pre-processing model 300 may include a function for calculating a vector of the input sentence itself based on the weight of the input sentence. The detailed description of the aforementioned pre-processing model 300 is merely an example, and the present disclosure is not limited thereto.

The processor 120 may obtain one or more reference sentence vectors by calculating each of the one or more reference sentences through the pre-processing model 300.

The reference sentence may be a sentence representing at least one of categories and attributes for classifying the user's input sentence. In addition, the reference sentence may be pre-stored in the database or the memory 130. And, each reference sentence may include a category and one or more attributes. For example, a reference sentence “What is a loan agreement?” May be stored in advance on the memory 130 as an attribute of “definition” in the “Terms & Conditions loan” category. In addition, the reference sentence "What are the terms and conditions of the loan?" May be stored in association with the reference sentence vector. In another embodiment, the reference sentence "What is the drugstore?" Is stored in the memory 130 and may be calculated as a reference sentence vector through the pre-processing model 300. In addition, a reference sentence with the "Contact" attribute for the "Insurance" category may include "I want to ask about insurance", "I want to know some insurance" and "I want to inquire about insurance" respectively. The detailed description of the above-mentioned reference sentence is only an example, and the present disclosure is not limited thereto.

The processor 120 may determine a reference sentence in which the similarity between each of the input sentence vector and the one or more reference sentence vectors is greater than or equal to a threshold value. The processor 120 may output an input sentence vector and one or more scoring values indicating the degree of similarity to each of the one or more reference sentence vectors. The processor 120 may determine a reference sentence based on a class of scoring values of a threshold value or more among the one or more scoring values. The processor 120 may calculate similarity by performing analysis based on the meaning, arrangement, and relationship between each word of one or more words included in the input sentence and the reference sentence. The processor 120 may calculate the similarity between the input sentence converted into the input sentence vector form in the preprocessing model 300 and the reference sentence vector stored in advance. Here, the similarity may be a score form scored in units of sentences.

For example, the processor 120 may analyze the similarity between the input sentences "I want to inquire about the terms of the loan" and "I want the terms of the loan" which are the reference sentences for the "Contact" property of the "Terms of Terms" category. Then, the processor 120 based on the matching of the terms "term" and "loan" included in the input sentence and the reference sentence, and the semantic similarity between "inquiry" and "want", the input sentence vector and the reference sentence vector The similarity of can be calculated as "80". In addition, the processor 120 may analyze the similarity between the input sentences "I want to inquire about the terms and conditions of the loan" and "Recommend the insurance products", which are reference sentences for the "Recommended" attribute of the "Insurance Products" category. Then, the processor 120 analyzes the degree to which the relationship between the terms & conditions and "inquiry" included in the input sentence matches the relationship between "insurance" and "recommended" included in the reference sentence, and the input sentence vector and The similarity with the reference sentence vector can be calculated as "35". The detailed operation description and numerical values of the above-described processor 120 are merely examples, and the present disclosure is not limited thereto.

In another example, the input sentence is “Are fractures guaranteed in my product?”, And the first reference sentence for the “coverage” attribute of the “insurance product” category is “What is the coverage information for the insurance product you subscribed to?” And, the second standard sentence may be "Can I claim insurance for this too?" In this case, the processor 120 may generate first and second similarities by calculating similarities between the input sentence vector and the first and second reference sentence vectors, respectively. More specifically, the processor 120 is based on the semantic similarity of the "I'm the product" and "Guarantee" of the input sentence and the "Subscribed insurance product" and the "Guarantee information" of the first reference sentence, and The first similarity of the 1 reference sentence vector can be calculated as “95”. And the processor 120 based on the semantic similarity of the "subscribed product" and "guaranteed information" of the input sentence and "insurance claim" of the second reference sentence, the second similarity of the input sentence vector and the second reference sentence vector It can be calculated as "61". Accordingly, the processor 120 may determine the first reference sentence based on the first similarity as the reference sentence based on the fact that the first similarity is greater than the second similarity. The above-described operation of the processor 120 is merely an example, and the present disclosure is not limited thereto.

In addition, when the similarity of at least one of the one or more reference sentences is equal to or greater than a threshold value, the processor 120 may transmit a response response sentence corresponding to the reference question sentence to the user terminal 200. More specifically, the processor 120 may calculate the similarity between the input sentence vector and one or more reference sentence vectors, respectively. Then, the processor 120 may compare the predetermined threshold value with the one or more similarities, respectively. Accordingly, the processor 120 may select a reference sentence having similarity or higher than a threshold value. Then, the processor 120 may select one of the reference sentences having a similarity level greater than or equal to the threshold value, and determine to transmit the response response stored in connection with the reference sentence to the user terminal 200. Response The response may include a response to an input sentence received from the user terminal. Here, the method in which the processor 120 selects one of the reference questions for which the similarity of the threshold value or more is calculated may be a method of selecting the highest similarity according to an embodiment.

In addition, the processor 120 according to an embodiment of the present disclosure may match the input sentence with the reference sentence having the largest excess value of the similarity compared to the threshold value when two or more reference sentences having a similarity threshold or higher. For example, the similarity to the first reference sentence may be “70”, and the similarity to the second reference sentence may be “80”. Also, the threshold value calculated by the processor 120 may be “68”. In this case, the processor 120 may classify the input sentence as the second reference sentence based on the fact that the threshold value of the similarity to the threshold value of the second reference sentence is greater than the first reference sentence. The above-described operation and numerical values of the processor 120 are merely examples, and the present disclosure is not limited thereto.

The processor 120 may transmit a response response to the user terminal 200 through the network unit 110 based on the reference sentence.

According to another embodiment of the present disclosure, a reference sentence may be determined based on an input sentence and a past input sentence received from the user terminal 200.

The processor 120 may obtain an input sentence vector by calculating an input sentence received from the user terminal 200 through the network unit 110 through a pre-processing model 300 including one or more network functions. When it is determined that it is difficult to determine the reference sentence using only the input sentence, the processor 120 may determine the reference sentence by calculating the past input sentence together with the input sentence. When it is determined that it is difficult to determine the reference sentence only with the input sentence, a scoring value equal to or greater than a threshold value among the scoring values obtained by determining the similarity between the reference sentence vector and the value converted into a vector form through the pre-processing model 300 for the input sentence This may include, for example, no case.

The processor 120 may obtain a past input sentence vector by calculating the past input sentence input before the input sentence received from the user terminal 200 through the network unit 110 through the pre-processing model 300. The processor 120 may determine a cumulative input vector based on the input sentence vector and the past input sentence vector. The cumulative input vector may be a sum vector or an average vector of input sentence vectors and past input sentence vectors.

For example, when the input sentence received by the processor 120 from the user terminal 200 through the network unit 110 is "What is the interest rate?", Each of two or more reference sentences including the "inquiry" property And at least some similarity, but the categories of each of the two or more reference sentences including the “inquiry” attribute are different (eg, “Term Loan” category, “Credit Loan” category) and compared with the entire reference sentence. In this case, there may be no class (ie, a reference sentence) in which the scoring value for similarity is greater than or equal to a threshold value. Accordingly, the processor 120 may determine the reference sentence based on the past input sentence and the input sentence received before the input sentence from the user terminal 200 through the network unit 110. For example, when the input sentence received by the processor 120 from the user terminal 200 is "What is the interest rate?", The vector for the past input sentence "Introduce about the terms and conditions loan" for the input sentence You can compute with vectors. For example, the processor 120 may obtain an input sentence vector for the input sentence “What is the interest rate?” And a past input sentence vector for the past input sentence “Introduce about drugs and loans”. For example, the processor 120 may determine a cumulative input vector by adding each input sentence vector and the past input sentence vector. The detailed description of the above-described input vector and past input vector is only an example, and the present disclosure is not limited thereto.

According to another embodiment of the present disclosure, when the processor 120 determines that it is difficult to determine a reference vector having a similarity or higher threshold value based on an input sentence received from the user terminal 200 through the network unit 110 , A reference sentence may be determined by vectorizing the past input sentence received from the user terminal 200 together with the input sentence. For example, if the input sentence received by the processor 120 is "What is the interest rate?", It can be calculated as a vector by combining with the sentence of the past input sentence, "Introduce about the loan agreement." For example, the processor 120 may convert a sentence about "what is the interest rate for introducing loan terms" to a vector to determine a cumulative input vector. The detailed description of the above-mentioned input sentence and past input sentence is only examples, and the present disclosure is not limited thereto.

 The processor 120 may determine the reference sentence whose degree of similarity with the cumulative input vector is greater than or equal to a threshold value. The processor 120 may determine the similarity between the input sentence converted into the cumulative input vector form in the preprocessing model 300 and the past input sentence, and the reference sentence vector stored in the dictionary, thereby determining a reference sentence having a similarity or higher than a threshold value. .

The processor 120 may determine that a reference sentence cannot be determined when there is no class corresponding to a scoring value of a threshold value or more among the one or more scoring values. The processor 120 may compare the similarity and threshold values for one or more reference sentences, and if all the similarities are less than the threshold value, it may determine that the input sentence is not classified as a reference question sentence. More specifically, the processor 120 may calculate the similarity by comparing the input sentence vector with each reference sentence vector. Then, the processor 120 may determine that the input sentence vector and all reference sentence vectors are below a threshold value.

The processor 120 may use at least one of an answer to determine whether a reference sentence based on a class of the maximum scoring value among the one or more scoring values is correct through the network unit 110 and a non-determinable answer to the input sentence. It can be transmitted to (200).

When there is no reference sentence vector above the threshold, the processor 120 may determine that the input sentence is not classified as the reference sentence. The processor 120 may determine, through the network unit 110, to transmit an answer including an apology message (eg, "sorry, but the question is still being learned") to the user terminal 200. For example, the threshold value preset by the processor 120 is “65”, and the network unit 110 may receive an input sentence “Are you bored?”. In this case, the processor 120 may convert the input sentence into an input sentence vector, and calculate similarities to all reference question sentence vectors as "14", "16", "22", and "20". In addition, the processor 120 may determine that the input sentence is not classified as a reference question sentence based on the point that all the similarities are less than the threshold value of “65”. The above-described operation of the processor 120 is merely an example, and the present disclosure is not limited thereto.

If there is no reference sentence vector above the threshold value, the processor 120 may provide an answer (eg, in the "account registration guide") to determine whether the reference sentence based on the class of the maximum scoring value among one or more class-specific scoring values below the threshold value is correct. Would you like to inquire about the content? "Or" Please select about the inquiry you made from the account registration guide and deposit account change method. ") To the user terminal 200. For example, the threshold value preset by the processor 120 is “65”, and the network unit 110 may receive an input sentence “I don't know what the abbreviation is”. In this case, the processor 120 may convert the input sentence into an input sentence vector, and calculate similarities with all reference question sentence vectors as "14", "63", "22", and "35". The processor 120 determines whether a reference sentence based on "63" having a maximum score value ("What is the drug?") Is similar to the similarity with one or more reference sentence vectors that are below a predetermined threshold value. An answer to confirm whether ("Which did you ask about?") May be transmitted to the user terminal 200 through the network unit 110. The above-described operation of the processor 120 is merely an example, and the present disclosure is not limited thereto.

The description of the above specific answer is only an example, and the present disclosure is not limited thereto.

The processor 120 may determine the topic of the conversation based on the sub-subject of the reference sentence.

In one embodiment of the present disclosure, the conversation may refer to a context of a sentence exchanged between the user and the computing device of the present disclosure. The conversation may include at least one sentence of an input sentence received from the user terminal and an answer transmitted to the user terminal.

The topic of the conversation may be a central issue in the chatbot continuing the conversation with the user. One subject may include one or more sub-themes. The sub topic may be a sub-problem belonging to a central problem in continuing a conversation with a user. One sub-subject may include one or more reference sentences. As described above, the reference sentence may be a sentence representing at least one of a category and an attribute for classifying a user's input sentence.

Hereinafter, the subject, sub-subject, and reference sentence will be described with reference to FIG. 3.

3 is a diagram illustrating subject and reference sentences stored in a server according to an embodiment of the present disclosure.

For example, the credit loan guide 310 and the terms loan guide 320 may each be a subject. The subject of the terms loan guide 320 may include one or more sub-themes. The subject of the terms loan guide 320 may include one or more sub-themes. The subject of the terms and conditions loan guide (320) is the sub-topics on the definition of terms and conditions loan (321), terms and conditions loan interest rate guide (322), login request (323), actual applied interest rate guide (324) and terms and conditions loan execution (325). Each can be included. When the processor 120 provides an answer to the user terminal 200 on the subject of the terms and conditions loan guidance 320, the processor 120 answers based on the order of the sub-themes (that is, the definition of terms and conditions loan (321)-terms and conditions loan Interest rate guidance 322-login request 323-actual application interest rate guidance 324-terms and conditions loan execution (325) in order to provide an answer to the user terminal 200) can be provided.

 For example, among one or more sub-subjects belonging to the subject of the terms and conditions loan guide 320, the terms and conditions loan definition guide 321 may include one or more reference sentences 330. The sub-topics of the terms and conditions loan definition guide 321 may include a standard sentence 330 such as what is pharmacy? (331), what is a terms loan? (332), please tell me about an insurance contract loan (333). .

When the subject is determined for the user terminal 200, the processor 120 may proceed with a conversation (ie, answer) based on the subject. The processor 120 may provide an answer to the user terminal 200 based on the order of the sub-themes for the user.

The processor 120 may determine a reference sentence having similarity or higher than a threshold value based on the input sentence received from the user terminal 200. For example, when the input sentence received by the processor 120 from the user terminal 200 is “Please tell me what the insurance contract loan is”, the input sentence vector for the input sentence is obtained through the preprocessing model 300. can do. The processor 120 may determine a reference sentence having a similarity level higher than or equal to a threshold based on the similarity between the reference sentence vector and the input sentence vector for each of the reference sentences 330. In this example, the processor 120 is similar to the threshold of the similarity of the reference sentence vector of “Please tell me about the insurance contract loan (333)” among the one or more reference sentences with respect to the input sentence vector of “Please tell me what the insurance contract loan is” With that in mind, you can decide on the standard sentence for “Please tell me about the insurance contract loan (333)”.

The processor 120 may check which sub-topic of the one or more sub-subjects is included in the determined reference sentence “Please tell me about the insurance contract loan (333)”. The processor 120 may determine a topic of a conversation as a basis for providing an answer to the user terminal 200 based on the sub-subject to which the determined reference sentence “Please tell me about the insurance contract loan (333)” belongs. For example, the processor 120 may check the sub topic “Guidelines for Definition of Terms and Conditions 321” which includes a standard sentence “Please tell me about insurance contract loans (333)”. The processor 120 may check the subject “contract loan guide 320” including the sub subject “contract loan definition guide 321”. The detailed description of the above-mentioned reference sentence, sub-subject, and subject is merely examples, and the present disclosure is not limited thereto.

According to another embodiment of the present disclosure, the processor 120 may determine a topic of the conversation based on the sub-subject of the reference sentence and the sub-subject of the past input sentence.

The processor 120 may determine a sub-subject of the reference sentence. When determining that it is not easy to determine the topic of the conversation based on the sub-subject of the reference sentence, the processor 120 may further determine the topic of the conversation by further checking the sub-subject based on the past input sentence. When the sub-subject of the reference sentence can be included in more than one subject, the processor 120 may determine that it is not easy to determine the topic of the conversation based on the sub-subject of the reference sentence.

For example, referring to FIG. 3, when the input sentence received by the processor 120 from the user terminal 200 is “tell me the interest rate”, it is not easy to determine the topic of the conversation only with the received input sentence I can judge. The processor 120 may determine that the sub-subject containing the reference sentence based on the input sentence for “tell me the interest rate” received from the user terminal 200 is “interest rate guide”. The processor 120 may determine that the sub-themes for “interest rate guide” can be included in the topics for “credit loan guide 310” and “contract loan guide 320”, respectively.

Alternatively, the processor 120 may determine a topic of the conversation by additionally checking the sub-subject based on the past input sentence, if it can be determined as a sub-subject for each sub-subject included in two or more topics based on the reference sentence. .

Or, for example, referring to FIG. 3, the processor 120 may refer to the reference sentence based on the input sentence for “tell me the interest rate” received from the user terminal 200, “credit loan interest rate guide 312” and It can be judged that it can be included in each of the sub-subjects corresponding to the “Terms of Interest Loan Interest 322”.

 That is, when it is difficult to determine which subject can be included only with the input sentence received from the user terminal 200, it is possible to check the past input sentence received from the user terminal 200.

The processor 120 may determine a reference past input sentence having a similarity level with a threshold value higher than or equal to a past input sentence input before the input sentence. The past input sentence may be an input sentence received before the input sentence from the user terminal 200. The past input sentence may include the latest predetermined number of input sentences received before the input sentence from the user terminal 200. The reference past input sentence may be a reference sentence for the past input sentence. The processor 120 may determine a reference sentence for “tell me a credit loan,” which is a past input sentence received before the input sentence from the user terminal 200. The processor 120 may determine a sub-subject of the reference past input sentence. The processor 120 may determine a sub-subject for the reference sentence for “tell me a credit loan”, which is the input sentence of the past, as the “guide for defining a credit loan 311”.

The processor 120 may determine a topic of conversation based on a sub-subject of the reference sentence and the sub-subject of the past input sentence. The processor 120 may determine two or more topics that may include sub-themes of the reference sentence. The processor 120 may determine a topic including a sub-subject of the reference past input sentence among the determined two or more topics as a topic for the input sentence. In the above example, the processor 120 may include two or more topics (“credit loan guide 310” and “contract loan guide”) which may include sub-subjects of the reference sentence (in this example, “interest rate guide”). (320) ”), and from the user terminal 200, a topic that may include a sub-topic based on a reference past input sentence (“ Guide to Definition of Credit Loan 311 ”in this example) among the two or more topics. You can decide on the subject of the received input sentence. The detailed description of the above-described subject and sub-topics is only examples, and the present disclosure is not limited thereto.

The processor 120 may determine the topic of the conversation based on the sub-subject of the reference sentence and the user's status information.

The processor 120 may determine the subject of the conversation based on the user's status information when the reference sentence cannot determine the subject of the conversation. The user's status information may include information about the user's past usage history. The processor 120 may determine the topic of the conversation based on the user's status information stored in the memory 130. The processor 120 may determine two or more topics that may include the sub-topic based on the sub-topics that include the reference sentence, and may include data based on the user's status information among the two or more topics. , It is possible to determine the topic of the conversation for the user terminal 200.

 For example, when the input sentence received from the user terminal 200 through the network unit 110 by the processor 120 is “hello”, the sub-subject for the input sentence may be determined as “greeting message”. The processor 120 may not be able to determine a specific one subject including the sub-theme. Since the processor 120 is unable to determine whether a response based on a subject is to the user terminal 200 based on the sub-subject, it is impossible to determine a dialogue based on the user's status information based on the user terminal 200. You can decide on the subject. The processor 120 may determine to respond to the user terminal 200 based on the subject of “credit loan guide 310” when the user's status information includes a history of the user receiving a credit loan. have.

When the subject of the conversation cannot be determined only by the reference sentence, the processor 120 may determine the subject of the conversation based on the user's status information and other user's status information. The processor 120 may compare state information of a user stored in the memory 130 with state information of another user. When the user's status information has data similar to the status information of another user, a recent topic applied to another user terminal may be determined as a topic applied to the user terminal. Whether the user's state information and other user's state information are similar may be converted into a vector through the pre-processing model 300 and determined based on the similarity of the vector values. If the vector based on the user's status information and the vector based on the other user's status information have a score value indicating a similarity of a predetermined threshold or higher, the processor 120 may apply the subject applied to the other user to the user. have. For example, the status information of another user includes the subject of “real life insurance for newborns” and the subject of “parent life insurance”, and the user's status information of “real life insurance for newborns” If it includes, the subject to be applied for answering the user terminal 200 may be determined as “parental life insurance” based on the subject applied to the other user. The detailed description of the state information of the user and the state information of another user is only an example, and the present disclosure is not limited thereto.

When there are two or more topics obtained based on the sub-topics, the processor 120 may transmit an answer for receiving a selection input for one of the two or more topics to the user terminal.

The processor 120 may determine whether a sub-subject of the reference sentence determined based on the input sentence received from the user terminal 200 is included in two or more subjects through the network unit 110. When two or more topics are obtained with respect to one determined sub-subject, the processor 120 receives selection input for one of the two or more topics to determine which of the two or more topics fits the user's question The response for the user may be transmitted to the user terminal 200 through the network unit 110.

For example, when the input sentence received from the user terminal 200 is "Can I borrow?", The sub-subject for the input sentence may be "Guide to the definition of the loan." Sub-subjects corresponding to “loan definition guide” may be included in topics for “credit loan guide 310” and “term loan guide 320”, respectively. The processor 120 receives an answer for receiving a selection input for one of the topics of the “credit loan guide 310” and the “condition loan guide 320” to the user terminal 200 through the network unit 110. Can transmit. The answer may be at least one of text and content including content to select a topic for one of two or more topics in order to receive a selection input for one topic at a user terminal. For example, the answer may be “Please enter the desired loan among credit loans and terms and conditions loans.” Or, the content may include clickable credit loans and clickable terms loans. The specific description of the above-mentioned answer is only an example, and the present disclosure is not limited thereto.

The processor 120 may transmit an answer based on the reference sentence of the sub-subject having the highest priority among the one or more sub-subjects included in the subject to the user terminal 200 through the network unit 110 to the user terminal.

The processor 120 may identify a subject based on an input sentence received from the user terminal 200 through the network unit 110.

The processor 120 may compare the input sentence vector and the reference sentence vector obtained by calculating the input sentence through the pre-processing model 300 and determine that there is no reference sentence vector having a similarity level higher than or equal to a threshold. When the reference sentence vector cannot be determined, the processor 120 may check whether the subject can be identified based on the input sentence. The processor 120 may obtain a subject vector by calculating a subject through the pre-processing model 300. The processor 120 may compare the similarity between the input sentence vector and the subject vector, and determine the subject for the user terminal 200 when the similarity has a value equal to or greater than a predetermined threshold.

The processor 120 may generate an answer based on a sub-subject having the highest priority among one or more sub-subjects included in the subject. The processor 120 may transmit the answer to the user terminal 200 through the network unit 110.

For example, when the input sentence received from the user terminal through the network unit 110 by the processor 120 is “weak”, there may be no reference sentence vector having a similarity or higher than a threshold value based on the input sentence vector based on the input sentence. have. The processor 120 may compare the input sentence vector and the subject vector based on the input sentence for “weakness” to determine a theme (in this example, “contract loan guidance 320”) having a similarity or higher than a threshold value. The processor 120 answers based on the “Guidelines for Definition of Terms and Conditions 321,” which is the highest priority sub-category among one or more sub-topics included in the “Terms and Conditions of Contract 320” topic (in this example, “ Can you tell me what the pharmacy is? ”) To the user terminal 200. The specific description of the above-mentioned subject and answer is merely an example, and the present disclosure is not limited thereto.

The processor 120 may determine an answer based on the first sub-subject corresponding to the next sub-subject of the sub-subject of the reference sentence among the one or more sub-subjects included in the subject.

In the present disclosure, the conversation may include one or more topics. The subject may include one or more sub-themes. One sub-subject may have a sequential relationship with another sub-subject depending on the progress of the conversation. The next sub-subject may mean a sub-subject that can come to one sub-subject subordinate, depending on the progress of the conversation. For example, when the subject of the conversation is a loan guide, the sub-subject to the actual applied interest rate guide 324 may be the next sub-subject to the sub-subject of the login request 323. In the above example, the sub-subject of the login request 323 and the sub-subject of the actual applied interest rate guide 324 may be interpreted as having a precedent relationship with each other. The detailed description of the aforementioned sub-subject is only an example, and the present disclosure is not limited thereto.

 The first sub-subject may be a sub-subject corresponding to the next order of the sub-subject based on the input sentence received from the user terminal 200. Response The response may include a response to an input sentence received from the user terminal. The answer may include a response for deriving an input sentence desired to be received from the user terminal. The answer may be an answer generated based on one reference sentence among one or more reference sentences included in the first sub-subject in the processor 120. Referring to FIG. 3, for example, when the sub-subject of the reference sentence is the “subscription loan definition guide 321”, the processor 120 subsub-subject of the topic “subscription loan guide 320” including the sub-subject. Among them, the first sub-subject can be confirmed. The first sub-subject may be “subscription loan interest rate guide 322” corresponding to the next sub-subject of “subscription loan definition guide 321” which is a sub-subject of the reference sentence. The processor 120 may generate an answer based on one reference sentence among the one or more reference sentences included in the first sub-subject (in this example, “the contract loan interest rate guide 322”). The processor 120 may generate an answer “Do you wonder about the interest rate?” Based on the reference sentence of the first sub-subject.

 At least one of the response response and the response may be transmitted by the processor 120 to the user terminal 200 through the network unit 110. The processor 120 may receive one input sentence from the user terminal 200 through the network unit 110 and transmit a response response and an answer before receiving the next input sentence. For example, when the input sentence received by the processor 120 from the user terminal 200 through the network unit 110 is “I wonder what the terms and conditions are”, the processor 120 has similarity to the corresponding input sentence By determining the high standard sentence, answering the above standard sentence (in this example, “Term Loan is a loan within the premium paid by the customer, and is often loaned within 70% of the insurance cancellation refund of the customer's subscription) It is a system that can receive. ”) Processor 120, the sub-subject to the reference sentence included in the subject determined based on the reference sentence (in this example, "Terms and Conditions Loan 320") (in this example, "Guidelines for the Definition of Terms and Conditions 321") ) May generate an answer based on one or more reference sentences included in the first sub-subject (in this example, “Term Loan Interest Rate Guide 322”), which is a sub-subject in the following order. The processor 120 may determine an answer generated based on the first sub-subject, “Do you wonder about the interest rate?”. The processor 120 responds to the user terminal 200 through the network unit 110 in response to the response (in this example, “the contract loan is a loan within the premium paid by the customer, 70 of the insurance cancellation refund of the customer's subscription) You can borrow from time to time within the% range. ”) And answers (in this example,“ Are you curious about the interest rate? ”).

The processor 120 may transmit at least one of a response response and an answer to the user terminal 200 through the network unit 110. The response response and the response may be transmitted by the processor 120 to the user terminal 200 through the network unit 110 through one transaction, or by transmitting two transactions respectively, to the display unit of the user terminal 200 Can be displayed. For example, the processor 120 is a transaction that includes a response and an answer to the user terminal 200 (in this example, “the contract loan is a loan within the premium paid by the customer, canceling the insurance that the customer subscribed to) This is a system that allows you to borrow from time to time within 70% of the refund amount. Do you have any questions about the interest rate? ”). In addition, for example, the processor 120 responds to the response of the response to the user terminal 200 and the response of each transaction (in this example, the first transaction, “Terms of Loan, means that the customer receives a loan within the premium paid by the customer,” This is a system that allows you to borrow at any time within 70% of the insurance cancellation refund. ”And the second transaction,“ Are you curious about the interest rate? ” It is only a specific description of the response response and the above-described response, and the present disclosure is not limited thereto.

The processor 120 may compare the sub-subject and the first sub-subject of the past input sentence to determine whether or not a conversation has been previously conducted with the user on a topic to be answered to the user. The processor 120 responds to the user's terminal 200 based on a second sub-subject that corresponds to a sequence subsequent to the first sub-subject when there is data in a conversation with the user about the subject to be answered to the user. On the network 110 can be transmitted.

The processor 120 may compare the first sub-subject and the sub-subject of the past input sentence input before the input sentence from the user terminal. When the first sub-subject and the sub-subject of the past input sentence match, the processor 120 may determine an answer based on the second sub-subject corresponding to the next sub-subject of the first sub-subject. The second sub-subject may be a sub-subject included in the same subject as the first sub-subject. The second sub-subject may be a sub-subject next to the first sub-subject. The second sub-subject may be a sub-subject of one or more sub-subjects in the next order of the first sub-subject.

Referring to FIG. 3, when the sub-subject determined based on the input sentence is the “term loan interest rate guide 322”, the processor 120 may include the sub-subject in the next order included in the same subject as the sub-subject determined above. An answer may be determined based on one sub-subject (in this example, “login request 323”). However, the processor 120 is a sub-subject of the past input sentence determined based on the past input sentence received from the user terminal 200 (in this example, “My ID is A and the password is B”) (in this example, “ If the login request 323 ”) is the same as the first sub-subject (“ login request 323 ”in this example), that is, it is determined that the user has already logged in before asking about the terms loan, 1 sub-subject ("login request 323" in this example) The answer may be determined based on the second sub-subject ("real-world application interest rate guide 324" in this example) corresponding to the following order. The specific description of the method for determining the above-mentioned answer is merely an example, and the present disclosure is not limited thereto.

The processor 120 may transmit an answer to the user terminal 200 through the network unit 110. When the answer includes content, in order to display the chat window based on the input sentence and the answer and the content together on the user terminal, the processor 120 may include a first layer including a chat window through the network unit 110 and the The interface to the second layer including the content may be transmitted to the user terminal 200.

The first layer may include a chat window including a sentence input by the user through the user terminal 200 and an answer transmitted from the processor 120. The second layer may include content. The content may include information necessary for the user and the chatbot to proceed with the consultation. Content may include photos, pictures, text, and the like. For example, the content may include a login screen, a loan agreement, an insurance brochure, an insurance contract document, insurance terms, and the like.

The first layer and the second layer may be included in one screen in the display unit of the user terminal 200. In one embodiment of the present disclosure, the first layer and the second layer may be included in one screen without overlapping. In another embodiment of the present disclosure, the first layer and the second layer may be included in one screen in an overlapped state. When the first layer and the second layer overlap, the upper layer is displayed as a semi-transparent layer, so that the contents of the lower layer can be seen even in the overlapped portion. The second layer including the content may be an upper layer of the first layer including the chat window. The second layer may be a document including text or the like to deliver information for consultation to the user. The second layer may be a document including an input field for the user to fill out a necessary part in the consultation process.

For example, when the processor 120 transmits an answer to the user terminal 200 through the network unit 110 to display the brochure for explaining the insurance to the user on the user terminal 200, for the insurance brochure The second layer including the content is an upper layer of the first layer and may be displayed on at least a part of the display unit of the user terminal 200 as a semi-transparent layer. For another example, when requesting a user to log in to check the interest rate for the user, in order to display a login window for receiving user input for the user ID and password input fields in the user terminal 200, the processor 120 May transmit an answer based on the second layer including the login content to the user terminal 200 through the network unit 110. The second layer including the login content may be disposed so as not to overlap with the first layer. The detailed contents of the contents of the first layer and the second layer described above are merely examples, and the present disclosure is not limited thereto.

2 is a diagram illustrating a flowchart between a server and a user terminal according to an embodiment of the present disclosure.

The computing device 100 may receive the past input sentence 201 from the user terminal 200. The past input sentence 201 may mean a sentence received before the computing device 100 receives the input sentence 202 from the user terminal 200. The input sentence 202 may include a sentence that the computing device 100 receives from the user terminal 200.

The computing device 100 may receive an input sentence 202 from the user terminal 200. The input sentence may include a sentence received by the computing device 100 from the user terminal 200.

The computing device 100 may calculate the similarity for each of the one or more reference sentences by calculating the input sentence 202 received from the user terminal 200 through the pre-processing model 300 including one or more network functions. More specifically, the computing device 100 may convert the input sentence 202 into an input sentence vector form by calculating through the pre-processing model 300 including one or more network functions. The pre-processing model 300 may be calculated in the form of a weighting matrix as a function for calculating a distance between vectors converted in units of morphemes. In addition, the pre-processing model 300 may include a function for calculating a vector of the input sentence 202 itself based on the weight of the input sentence 202. The computing device 100 may transmit a response answer to the user terminal 200 based on the reference question sentence above the threshold value when the similarity to at least one of the one or more reference question sentences is above the threshold value 101. Response The response may include a response to the input sentence 202 received from the user terminal.

The computing device 100 may determine a reference sentence based on the input sentence 202 and the past input sentence 201 received from the user terminal 200. The computing device 100 calculates a past input sentence vector by calculating the past input sentence 201 input before the input sentence 202 received from the user terminal 200 through the network unit 110 through the pre-processing model 300. Can be obtained. The computing device 100 may determine a cumulative input vector based on the input sentence vector and the past input sentence vector. The cumulative input vector may be a sum vector or an average vector of input sentence vectors and past input sentence vectors. The computing device 100 calculates the similarity between the input sentence 202 and the past input sentence 201 converted in the form of a cumulative input vector in the preprocessing model 300, and the reference sentence vector stored in advance, so that the similarity is a threshold value. The above standard sentence can be determined. The computing device 100 may transmit a response response to the user terminal 200 based on the determined reference sentence.

The computing device 100 may check which sub-subject of the determined reference sentence is included in one or more sub-subjects. The computing device 100 may determine a topic of a conversation as a basis for providing an answer 102 to the user terminal 200 based on the sub-subject to which the determined reference sentence belongs.

The computing device 100 may determine the answer 102 based on the first sub-subject corresponding to the next sub-subject of the sub-subject of the reference sentence among the one or more sub-subjects included in the subject. The first sub-subject may be a sub-subject corresponding to the next order of the sub-subjects based on the input sentence 202 received from the user terminal 200. The answer 102 may include a response for deriving an input sentence desired to be received from the user terminal. The answer 102 may be an answer generated based on one reference sentence among one or more reference sentences included in the first sub-subject in the computing device 100.

The computing device 100 may transmit the answer 102 to the user terminal 200.

The computing device 100 does not have a scoring value that is equal to or greater than a threshold value 101 among the scores obtained by determining the similarity between the reference sentence vector and the value converted into a vector form through the pre-processing model 300 with respect to the input sentence 202. In this case, it can be determined that the reference sentence cannot be determined. The computing device 100 may transmit a non-determinable answer 103 for the input sentence 202 to the user terminal 200.

4 is a diagram illustrating an operation interface of a chatbot according to an embodiment of the present disclosure.

The computing device 100 may receive a request for establishing a session related to a chatbot from the user terminal 200. Then, the computing device 100 may establish a chat session between the chatbot and the user terminal 200 based on the request. Also, the computing device 100 may output an input sentence input by the user terminal 200 to the user terminal 200 or transmit an interface for outputting an answer.

The computing device 100 may receive the input sentence 411 "What is the drugstore?" From the user terminal 200. Here, the input sentence 411 is disposed adjacent to one side of the interface (eg, the left side as shown in FIG. 4), and a user terminal icon 410 that can identify the user terminal 200 may be tagged. . In addition, the processor 120 may input the received input sentence "What is a drugstore?" To the pre-processing model 300 and classify it as a "question" attribute of the "Terms and Conditions" category. In addition, the computing device 100 may generate a response answer 421 "Insurance contract loan (term loan) is ~" based on category and attribute classification. Detailed description related to the operation of the computing device 100 is omitted because it has been described above with reference to FIG. 1.

The computing device 100 transmits the generated response response 421 "Insurance contract loan (term loan) is ~" to the user terminal 200. Here, the response answer 421 is disposed adjacent to one side (eg, the right side as shown in FIG. 4) of the interface, and a chatbot icon 420 capable of identifying the chatbot may be tagged.

The computing device 100 may check which sub-subject of the determined reference sentence is included in one or more sub-subjects. The computing device 100 may determine a topic of a conversation as a basis for providing an answer to the user terminal 200 based on the sub-subject to which the determined reference sentence belongs.

The computing device 100 determines the answer 422 “Do you not wonder about the interest rate?” Based on the first sub-subject corresponding to the next sub-subject of the sub-subject of the reference sentence among the one or more sub-subjects included in the subject. You can. The first sub-subject may be a sub-subject corresponding to the next order of the sub-subject based on the input sentence 411 received from the user terminal 200. The response response 421 may include a response to the input sentence 411 received from the user terminal. The answer 422 may include a response for deriving an input sentence desired to be received from the user terminal. The answer 422 may be an answer generated based on one reference sentence among one or more reference sentences included in the first sub-subject in the computing device 100.

The computing device 100 transmits the generated answer 422 “Do you not wonder about the interest rate?” To the user terminal 200. Here, the answer 422 is disposed adjacent to one side (eg, the right side as shown in FIG. 4) of the interface, and a chatbot icon capable of identifying the chatbot may be tagged.

The computing device 100 may receive the input sentence 412 “tell me the interest rate” according to the answer 422 from the user terminal 200.

The computing device 100 may transmit the answer 423 “Please log in to check the interest rate” to the user terminal 200 based on the input sentence 412 received from the user terminal 200. The answer 423 may include content 424. The content 424 may include information necessary for the user and the chatbot to proceed with the consultation. Content may include photos, pictures, text, and the like. The answer 423 may include ID input content for login and password input content. The interface for providing the answer 423 to the user terminal 200 may include a first layer and a second layer. The first layer may include a chat window, and the second layer may include the content 424 of the answer 423.

The first layer including the chat window may be disposed over the entire interface. The second layer including the content may be disposed on at least a part of the interface (eg, at the bottom right as shown in FIG. 4).

5 is a flowchart of a method for driving a chatbot according to an embodiment of the present disclosure.

The computing device 100 may determine 510 a reference sentence having a similarity level to or higher than a threshold value from the input sentence received from the user terminal. The computing device 100 may obtain an input sentence vector by calculating an input sentence through a pre-processing model including one or more network functions. The computing device 100 may obtain one or more reference sentence vectors by calculating each of the one or more reference sentences through the pre-processing model. The computing device 100 may determine a reference sentence having a similarity level equal to or greater than a threshold value for each of the input sentence vector and the one or more reference sentence vectors. The computing device 100 may convert the input sentence into an input sentence vector form by calculating through the pre-processing model 300 including one or more network functions. Here, the pre-processing model 300 may include a function for converting an input sentence into a set of one or more morphemes through morpheme analysis. The pre-processing model 300 may be calculated in the form of a weighting matrix using a function for calculating a distance between vectors converted in morpheme units. In addition, the pre-processing model 300 may include a function for calculating a vector of the input sentence itself based on the weight of the input sentence.

The computing device 100 may obtain an input sentence vector by calculating an input sentence through a pre-processing model including one or more network functions. The computing device 100 may obtain a past input sentence vector by calculating a past input sentence input before the input sentence through the preprocessing model. The computing device 100 may determine a cumulative input vector based on the input sentence vector and the past input sentence vector. The computing device 100 may determine a reference sentence whose degree of similarity with the cumulative input vector is greater than or equal to a threshold value.

The computing device 100 may output an input sentence vector and one or more scoring values indicating the degree of similarity for each of the one or more reference sentence vectors. The computing device 100 may determine a reference sentence based on a class of scoring values of a threshold value or more among one or more scoring values. The computing device 100 may determine that the reference sentence cannot be determined when there is no class corresponding to a scoring value of a threshold value or more among the one or more scoring values. The computing device 100 may transmit at least one of an answer for determining whether the reference sentence based on the class of the maximum scoring value among the one or more scoring values is correct and an undecidable answer for the input sentence to the user terminal.

The computing device 100 may determine a sub-subject of the reference sentence and determine 520 a topic of the conversation based on the sub-subject of the reference sentence. The computing device 100 may determine a reference past input sentence that has a similarity to a past input sentence that is input before the input sentence or more than a threshold value. The computing device 100 may determine a sub-subject of the reference past input sentence. The computing device 100 may determine the subject of the conversation based on the sub-subject of the reference sentence and the sub-subject of the reference past input sentence.

The computing device 100 may transmit an answer for receiving a selection input for one of the two or more topics to the user terminal when there are two or more topics obtained based on the sub-themes.

The computing device 100 may determine the topic of the conversation based on the sub-subject of the reference sentence and the user's status information.

The computing device 100 may determine 530 an answer based on a first sub-subject corresponding to a next sub-subject of the sub-subject of the reference sentence among a plurality of sub-subjects included in the topic of the conversation. The answer may include a response for deriving an input sentence desired to be received from the user terminal.

The computing device 100 may compare the first sub-subject and the sub-subject of the past input sentence input before the input sentence from the user terminal. When the first sub-subject and the sub-subject of the past input sentence match, the computing device 100 may determine an answer based on the second sub-subject corresponding to the next sub-subject of the first sub-subject.

The computing device 100 may transmit the answer 540 to the user terminal.

When the answer includes the content, the computing device 100 may display a chat window based on the input sentence and the answer and the content together on the first terminal including the chat window and the second layer including the content. The interface can be transmitted to the user terminal.

In one embodiment of the present disclosure, the computing device 100 may determine at least one of a subject and a sub-subject of the user's input sentence according to a similarity comparison between a user's input sentence and a reference sentence, and at least one of the determined subject and sub-subjects Based on one, it is possible to provide answers for the next conversation with the user, so that a chatbot capable of leading the conversation with the user can be provided.

6 is a block diagram illustrating logic for implementing a method of driving a chatbot according to an embodiment of the present disclosure.

According to an embodiment of the present disclosure, a method of driving a chatbot may be implemented by the following logic.

Logic 610 for determining a reference sentence having a similarity or higher than a threshold value with an input sentence received from a user terminal according to an embodiment of the present disclosure; Logic 620 for determining a sub-subject of the reference sentence and determining a topic of a conversation based on the sub-subject of the reference sentence; A logic 630 for determining an answer based on a first sub-subject corresponding to a next sub-subject of the sub-subject of the reference sentence among a plurality of sub-subjects included in the topic of the conversation; And logic 640 for transmitting the answer to the user terminal.

In an embodiment of the present disclosure, logic 610 for determining a reference sentence having a similarity to an input sentence received from a user terminal that is greater than or equal to a threshold value is input by calculating the input sentence through a pre-processing model including one or more network functions. Logic to obtain a sentence vector; Logic for obtaining one or more reference sentence vectors by computing each of the one or more reference sentences through the pre-processing model; And logic for determining a reference sentence in which the similarity between each of the input sentence vector and the one or more reference sentence vectors is greater than or equal to a threshold value.

In an embodiment of the present disclosure, logic 610 for determining a reference sentence having a similarity to an input sentence received from a user terminal that is greater than or equal to a threshold value is input by calculating the input sentence through a pre-processing model including one or more network functions. Logic to obtain a sentence vector; Logic for obtaining a past input sentence vector by calculating a past input sentence input before the input sentence through the pre-processing model; Logic for determining a cumulative input vector based on the input sentence vector and the past input sentence vector; The similarity with the cumulative input vector may be implemented by logic for determining the reference sentence above a threshold value.

In an embodiment of the present disclosure, logic for determining a sub-subject of the reference sentence and determining a topic of a conversation based on the sub-subject of the reference sentence includes: logic for determining a sub-subject of the reference sentence; Logic for determining a reference past input sentence having a similarity level with a threshold value equal to or higher than a past input sentence input before the input sentence; Logic for determining a sub-subject of the reference past input sentence; And logic for determining a topic of conversation based on the sub-subject of the reference sentence and the sub-subject of the reference past input sentence.

In an embodiment of the present disclosure, logic 620 for determining a sub-subject of the reference sentence and determining a topic of a conversation based on the sub-subject of the reference sentence includes: a sub-subject of the reference sentence and user status information It can also be implemented by logic to determine the topic of a conversation based on it.

In an embodiment of the present disclosure, logic for determining a reference sentence having a similarity level of each of the input sentence vector and each of the one or more reference sentence vectors equal to or greater than a threshold value includes a degree of similarity between each of the input sentence vector and the one or more reference sentence vectors. Logic for outputting one or more scoring values indicative of; Logic for determining a reference sentence based on a class of scoring values equal to or greater than a threshold value among the one or more scoring values; And if there is no class corresponding to a scoring value of a threshold value or more among the one or more scoring values, it may be implemented by logic for determining that a reference sentence cannot be determined.

In an embodiment of the present disclosure, when there is no class corresponding to a scoring value of a threshold value or more among the one or more scoring values, logic for determining that a reference sentence cannot be determined is selected from the class of the maximum scoring value among the one or more scoring values. It may also be implemented by logic for transmitting at least one of an answer for determining whether the basic reference sentence is correct and an answer for which the input sentence cannot be determined to the user terminal.

In an embodiment of the present disclosure, the logic 640 for transmitting the answer to the user terminal displays, when the answer includes content, a chat window based on the input sentence and the answer and the content on the user terminal together In order to do so, the interface for the first layer including the chat window and the second layer including the content may be implemented by logic for transmitting to the user terminal.

In an embodiment of the present disclosure, a logic 630 for determining an answer based on a first sub-subject that corresponds to a next sub-subject of the sub-subject of the reference sentence among the plurality of sub-subjects included in the topic of the conversation, the Logic for comparing a first sub-subject and a sub-subject of a past input sentence input before the input sentence from the user terminal; And when the first sub-subject and the sub-subject of the past input sentence match, it may be implemented by logic for determining the answer based on the second sub-subject corresponding to the next sub-subject of the first sub-subject. .

In an embodiment of the present disclosure, when there are two or more topics obtained based on the sub-topics, it may be implemented by logic to transmit an answer for receiving a selection input for one of the two or more topics to the user terminal. have.

Logic for implementing a method for driving a chatbot according to an embodiment of the present disclosure may be implemented by a module, circuit, or means for implementing a method for driving a chatbot.

7 is a block diagram of a computing device according to an embodiment of the present disclosure.

7 shows a simplified and general schematic diagram of an exemplary computing environment in which embodiments of the present disclosure can be implemented.

Although the present disclosure has been described above in general with respect to computer-executable instructions that can be executed on one or more computers, those skilled in the art will appreciate that the present disclosure can be implemented in combination with other program modules and / or as a combination of hardware and software. will be.

Generally, program modules include routines, programs, components, data structures, etc. that perform particular tasks or implement particular abstract data types. In addition, those of ordinary skill in the art may find that the methods of the present disclosure include single-processor or multiprocessor computer systems, minicomputers, mainframe computers, as well as personal computers, handheld computing devices, microprocessor-based or programmable consumer electronics, and the like (each of which is). It will be appreciated that it may be implemented in other computer system configurations, including one that may operate in conjunction with one or more associated devices.

The described embodiments of the present disclosure may also be practiced in distributed computing environments where certain tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote memory storage devices.

Computers typically include a variety of computer readable media. Any computer-accessible medium can be a computer-readable medium. Computer-readable media includes volatile and non-volatile media, transitory and non-transitory media, removable and non-removable media. By way of example, and not limitation, computer readable media may include computer readable storage media and computer readable transmission media. Computer-readable storage media are volatile and non-volatile media, temporary and non-transitory media, removable and non-removable media implemented in any method or technology for storing information such as computer readable instructions, data structures, program modules or other data Includes media. Computer storage media include RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital video disk (DVD) or other optical disk storage devices, magnetic cassettes, magnetic tapes, magnetic disk storage devices or other magnetic storage devices, Or any other medium that can be accessed by a computer and used to store desired information.

Computer readable transmission media typically embodies computer readable instructions, data structures, program modules or other data in a modulated data signal, such as a carrier wave or other transport mechanism, and Includes all information delivery media. The term modulated data signal means a signal in which one or more of the characteristics of the signal are set or changed to encode information in the signal. By way of example, and not limitation, computer readable transmission media includes wired media such as a wired network or direct-wired connection, and wireless media such as acoustic, RF, infrared, and other wireless media. Combinations of any of the above-described media are also intended to be included within the scope of computer-readable transmission media.

An exemplary environment 1100 is shown that implements various aspects of the present disclosure, including a computer 1102, and the computer 1102 includes a processing device 1104, a system memory 1106, and a system bus 1108. do. System bus 1108 connects system components, including, but not limited to, system memory 1106 to processing device 1104. The processing device 1104 can be any of a variety of commercial processors. Dual processor and other multiprocessor architectures may also be used as the processing unit 1104.

The system bus 1108 can be any of several types of bus structures that can be further interconnected to a memory bus, peripheral bus, and local bus using any of a variety of commercial bus architectures. System memory 1106 includes read-only memory (ROM) 1110 and random access memory (RAM) 1112. The basic input / output system (BIOS) is stored in a non-volatile memory 1110 such as ROM, EPROM, EEPROM, etc. This BIOS is a basic to help transfer information between components in the computer 1102 at the same time as during startup. Contains routines. The RAM 1112 may also include high-speed RAM, such as static RAM for caching data.

Computer 1102 is also an internal hard disk drive (HDD) 1114 (eg, EIDE, SATA)-this internal hard disk drive 1114 can also be configured for external use within a suitable chassis (not shown). Yes, magnetic floppy disk drive (FDD) 1116 (e.g., for reading from or writing to removable diskette 1118), and optical disk drive 1120 (e.g., CD-ROM Disc 1122, or for reading from or writing to other high-capacity optical media such as a DVD). The hard disk drive 1114, the magnetic disk drive 1116, and the optical disk drive 1120 are respectively connected to the system bus 1108 by the hard disk drive interface 1124, the magnetic disk drive interface 1126, and the optical drive interface 1128. ). The interface 1124 for implementing an external drive includes at least one or both of Universal Serial Bus (USB) and IEEE 1394 interface technologies.

These drives and their associated computer-readable media provide non-volatile storage of data, data structures, computer-executable instructions, and the like. In the case of computer 1102, drives and media correspond to storing any data in a suitable digital format. Although the above description of computer readable media refers to HDDs, removable magnetic disks, and removable optical media such as CDs or DVDs, those of ordinary skill in the art can use a zip drive, magnetic cassette, flash memory card, cartridge, etc. It will be appreciated that other types of media readable by a computer, etc., may also be used in the exemplary operating environment and any such media may include computer-executable instructions for performing the methods of the present disclosure.

A number of program modules may be stored in the drive and RAM 1112, including the operating system 1130, one or more application programs 1132, other program modules 1134, and program data 1136. All or part of the operating system, applications, modules and / or data may also be cached in RAM 1112. It will be appreciated that the present disclosure can be implemented in various commercially available operating systems or combinations of operating systems.

The user may input commands and information to the computer 1102 through one or more wired / wireless input devices, for example, pointing devices such as a keyboard 1138 and a mouse 1140. Other input devices (not shown) may include a microphone, IR remote control, joystick, game pad, stylus pen, touch screen, etc. These and other input devices are often connected to the processing unit 1104 through an input device interface 1142 connected to the system bus 1108, but the parallel port, IEEE 1394 serial port, game port, USB port, IR interface, And other interfaces.

The monitor 1144 or other type of display device is also connected to the system bus 1108 through an interface such as a video adapter 1146. In addition to the monitor 1144, the computer generally includes other peripheral output devices (not shown) such as speakers, printers, and the like.

The computer 1102 can operate in a networked environment using logical connections to one or more remote computers, such as remote computer (s) 1148 via wired and / or wireless communication. The remote computer (s) 1148 may be a workstation, computing device computer, router, personal computer, portable computer, microprocessor-based entertainment device, peer device, or other conventional network node, typically in computer 1102. It includes many or all of the components described with respect to, but for simplicity, only the memory storage device 1150 is shown. The illustrated logical connections include wired / wireless connections to a local area network (LAN) 1152 and / or a larger network, such as a wide area network (WAN) 1154. Such LAN and WAN networking environments are common in offices and companies, and facilitate enterprise-wide computer networks such as intranets, all of which can be connected to computer networks around the world, for example, the Internet.

When used in a LAN networking environment, the computer 1102 is connected to the local network 1152 via a wired and / or wireless communication network interface or adapter 1156. The adapter 1156 may facilitate wired or wireless communication to the LAN 1152, which also includes a wireless access point installed therein to communicate with the wireless adapter 1156. When used in a WAN networking environment, the computer 1102 may include a modem 1158, connect to a communication computing device on the WAN 1154, or establish communication over the WAN 1154, such as through the Internet. Have other means. The modem 1158, which may be an internal or external and wired or wireless device, is connected to the system bus 1108 through a serial port interface 1142. In a networked environment, program modules described for the computer 1102 or portions thereof may be stored in the remote memory / storage device 1150. It will be appreciated that the network connections shown are exemplary and other means of establishing a communication link between computers may be used.

The computer 1102 is associated with any wireless device or object that is deployed and operates in wireless communication, such as a printer, scanner, desktop and / or portable computer, a portable data assistant (PDA), communication satellite, or wireless detectable tag. It operates to communicate with any equipment or place and telephone. This includes at least Wi-Fi and Bluetooth wireless technology. Accordingly, the communication may be a predefined structure as in a conventional network or simply ad hoc communication between at least two devices.

Wi-Fi (Wireless Fidelity) enables a connection to the Internet or the like without a wired connection. Wi-Fi is a wireless technology such as a cell phone that allows a device, for example, a computer to transmit and receive data indoors and outdoors, ie anywhere within the base station's coverage area. Wi-Fi networks use a wireless technology called IEEE 802.11 (a, b, g, etc.) to provide a secure, reliable and high-speed wireless connection. Wi-Fi can be used to connect computers to each other, to the Internet, and to a wired network (using IEEE 802.3 or Ethernet). Wi-Fi networks can operate in unlicensed 2.4 and 5 GHz radio bands, for example, at 11 Mbps (802.11a) or 54 Mbps (802.11b) data rates, or in products that include both bands (dual band). have.

One of ordinary skill in the art of the present disclosure will understand that information and signals can be represented using any of a variety of different technologies and techniques. For example, data, instructions, instructions, information, signals, bits, symbols and chips that can be referenced in the above description are voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields Or particles, or any combination thereof.

Those of ordinary skill in the art of the present disclosure may use various examples of logical blocks, modules, processors, means, circuits, and algorithm steps described in connection with the embodiments disclosed herein in electronic hardware, (convenience For the sake of understanding, it may be implemented by various forms of program or design code (referred to herein as “software”) or a combination of both. To clearly illustrate this interoperability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the particular application and design constraints imposed on the overall system. Those skilled in the art of the present disclosure may implement the functions described in various ways for each particular application, but such implementation decisions should not be interpreted as being outside the scope of the present disclosure.

Various embodiments presented herein may be implemented as a method, apparatus, or article of manufacture using standard programming and / or engineering techniques. The term "manufactured article" includes a computer program, carrier, or media accessible from any computer-readable device. For example, computer-readable media include magnetic storage devices (eg, hard disks, floppy disks, magnetic strips, etc.), optical disks (eg, CDs, DVDs, etc.), smart cards, and flash memory Devices (eg, EEPROMs, cards, sticks, key drives, etc.), but are not limited to these. Also, various storage media presented herein include one or more devices and / or other machine-readable media for storing information.

It is understood that the specific order or hierarchy of steps in the processes presented is an example of exemplary approaches. Based on design priorities, it is understood that within the scope of the present disclosure, a specific order or hierarchy of steps in the processes may be rearranged. The accompanying method claims provide elements of the various steps in a sample order, but are not meant to be limited to the specific order or hierarchy presented.

Descriptions of the presented embodiments are provided to enable any person of ordinary skill in the art to use or practice the present disclosure. Various modifications to these embodiments will be apparent to those skilled in the art of the present disclosure. The general principles defined herein can be applied to other embodiments without departing from the scope of the present disclosure. Thus, the present disclosure should not be limited to the embodiments presented herein, but should be interpreted in the broadest scope consistent with the principles and novel features presented herein.

Claims (13)

A computer program stored in a computer-readable storage medium, wherein the computer program, when executed on one or more processors, performs the following operations for a method of driving a chatbot, the operations being:
Determining a reference sentence having a similarity level or higher with an input sentence received from the user terminal;
Determining a sub-subject of the reference sentence, and determining whether the sub-subject is included in a topic of two or more conversations;
When the sub-subject is included in the subject of one conversation, determining a subject including the sub-subject as the subject of the conversation;
When the sub-subject is included in a topic of two or more conversations, a reference past input sentence having similarity with a past input sentence input before the input sentence is greater than or equal to a threshold value is determined, and a sub-subject of the reference past input sentence is determined, and the Determine a topic of the conversation based on a sub-subject of the reference sentence and the sub-subject of the reference past input sentence, or receive an answer for receiving a selection input for the topic of one of the two or more conversation topics Transmitting to a user terminal;
Determining an answer based on a first sub-subject corresponding to a next sub-subject of the sub-subject of the reference sentence among a plurality of sub-subjects included in the topic of the conversation; And
Transmitting the answer to the user terminal;
Containing,
A computer program stored on a computer readable storage medium.
According to claim 1,
The answer includes a response for deriving an input sentence desired to be received from the user terminal,
A computer program stored on a computer readable storage medium.
According to claim 1,
Determining a reference sentence having a similarity or higher than a threshold value with the input sentence received from the user terminal,
Obtaining an input sentence vector by computing the input sentence through a pre-processing model including one or more network functions;
Obtaining one or more reference sentence vectors by calculating each of the one or more reference sentences through the pre-processing model; And
Determining a reference sentence having a similarity level equal to or greater than a threshold value for each of the input sentence vector and the one or more reference sentence vectors;
Containing,
A computer program stored on a computer readable storage medium.
According to claim 1,
Determining a reference sentence having a similarity or higher than a threshold value with the input sentence received from the user terminal,
Obtaining an input sentence vector by computing the input sentence through a pre-processing model including one or more network functions;
Obtaining a past input sentence vector by calculating a past input sentence input before the input sentence through the pre-processing model;
Determining a cumulative input vector based on the input sentence vector and the past input sentence vector; And
Determining the reference sentence having a degree of similarity with the accumulated input vector equal to or greater than a threshold value;
Containing,
A computer program stored on a computer readable storage medium.
delete According to claim 1,
When the sub-subject is included in a topic of two or more conversations, a reference past input sentence having similarity with a past input sentence input before the input sentence is greater than or equal to a threshold value is determined, and a sub-subject of the reference past input sentence is determined, and the Determine a topic of the conversation based on a sub-subject of the reference sentence and the sub-subject of the reference past input sentence, or receive an answer for receiving a selection input for the topic of one of the two or more conversation topics The operation to transmit to the user terminal,
Determining a topic of the conversation based on the sub-subject of the reference sentence and status information of a user;
Containing,
A computer program stored on a computer readable storage medium.
The method of claim 3,
Determining a reference sentence having a similarity level equal to or greater than a threshold value for each of the input sentence vector and the one or more reference sentence vectors includes:
Outputting one or more scoring values indicating a degree of similarity to each of the input sentence vector and the one or more reference sentence vectors;
Determining a reference sentence based on a class of scoring values equal to or greater than a threshold value among the one or more scoring values; And
Determining that a reference sentence cannot be determined when there is no class corresponding to a scoring value of a threshold value or more among the one or more scoring values;
Containing,
A computer program stored on a computer readable storage medium.
The method of claim 7,
If there is no class corresponding to a scoring value of a threshold value or more among the one or more scoring values, the operation of determining that the reference sentence cannot be determined is:
Transmitting at least one of an answer to determine whether a reference sentence based on a class of the maximum scoring value among the one or more scoring values is correct and an answer that cannot be determined for the input sentence to the user terminal;
Containing,
A computer program stored on a computer readable storage medium.
According to claim 1,
The operation of transmitting the answer to the user terminal,
If the answer includes content,
To display the chat window and the content based on the input sentence and the answer together on the user terminal,
Transmitting an interface for a first layer including the chat window and a second layer including the content to the user terminal;
Containing,
A computer program stored on a computer readable storage medium.
According to claim 1,
The operation of determining an answer based on a first sub-subject corresponding to the next sub-subject of the sub-subject of the reference sentence among the plurality of sub-subjects included in the topic of the conversation,
Comparing the first sub-subject and the sub-subject of a past input sentence input before the input sentence from the user terminal; And
If the first sub-subject and the sub-subject of the past input sentence match, determining the answer based on a second sub-subject corresponding to the next sub-subject of the first sub-subject;
Containing,
A computer program stored on a computer readable storage medium.
delete A method for driving a chatbot performed on one or more processors of a computing device,
Determining, by the processor, a reference sentence having a similarity level or higher with a threshold value that is received from a user terminal through a network unit;
Determining, by the processor, a sub-subject of the reference sentence, and determining whether the sub-subject is included in a topic of two or more conversations;
When the sub-subject is included in the subject of one conversation in the processor, determining a subject including the sub-subject as the subject of the conversation;
When the sub-subject is included in the subject of two or more conversations in the processor, a reference past input sentence having similarity to a past input sentence input before the input sentence is greater than or equal to a threshold value is determined, and a sub-subject of the reference past input sentence is determined. To determine the subject of the conversation based on the sub-subject of the reference sentence and the sub-subject of the reference past input sentence, or to receive a selection input for the subject of one of the two or more conversation subjects Transmitting an answer to the user terminal through the network unit;
Determining, by the processor, an answer based on a first sub-subject corresponding to a next sub-subject of the sub-subject of the reference sentence among a plurality of sub-subjects included in the topic of the conversation; And
Transmitting the answer from the processor to the user terminal through the network unit;
Containing,
A method for driving a chatbot performed on one or more processors of a computing device.
As a computing device for driving the chatbot,
A processor comprising one or more cores;
Network unit; And
Memory;
Including,
The processor,
Determine a reference sentence having a similarity or higher than a threshold value with the input sentence received from the user terminal,
Determine a sub-subject of the reference sentence, determine whether the sub-subject is included in a topic of two or more conversations,
When the sub-subject is included in the subject of one conversation, a subject including the sub-subject is determined as the subject of the conversation,
When the sub-subject is included in a topic of two or more conversations, a reference past input sentence having similarity with a past input sentence input before the input sentence is greater than or equal to a threshold value is determined, and a sub-subject of the reference past input sentence is determined, and the Determine a topic of the conversation based on a sub-subject of the reference sentence and the sub-subject of the reference past input sentence, or receive an answer for receiving a selection input for the topic of one of the two or more conversation topics Transmitted to the user terminal,
A response is determined based on a first sub-subject corresponding to a next sub-subject of the sub-subject of the reference sentence among a plurality of sub-subjects included in the topic of the conversation, and
Transmitting the answer to the user terminal,
A computing device for driving the chatbot.

Images