WO2019147039A1

WO2019147039A1 - Method for determining optimal conversation pattern for goal achievement at particular time point during conversation session associated with conversation understanding ai service system, method for determining goal achievement prediction probability, and computer-readable recording medium

Info

Publication number: WO2019147039A1
Application number: PCT/KR2019/001015
Authority: WO
Inventors: 설재호; 장세영
Original assignee: 주식회사 머니브레인
Priority date: 2018-01-26
Filing date: 2019-01-24
Publication date: 2019-08-01
Also published as: KR101945983B1

Abstract

The present disclosure provides a method for, in order to achieve a predetermined goal, determining a goal achievement prediction probability at a particular time point in process of a conversation session between one user terminal among multiple user terminals and a conversation understanding AI service server on a conversation understanding AI service server configured to process a natural language conversation for each of the multiple user terminals. A method according to the present disclosure comprises the steps of: receiving a natural language input from a user terminal at a particular time point; determining an intent corresponding to the received natural language input; determining a conversation history at the particular time point, wherein the conversation history indicates a flow of intents between the user terminal and a conversation understanding AI service server generated to the particular time point in a conversation session and includes an intent determined at the end of the flow of the intents; determining at least one conversation pattern including the conversation history among conversation patterns on the basis of the conversation history and a template prepared in advance for the conversation understanding AI service server, wherein the template includes multiple conversation patterns, each of which represents a flow of corresponding intents, and each conversation pattern has the number of times of goal achievement success and the number of times of goal achievement failure; and determining a goal achievement prediction probability at the particular time point on the basis of the corresponding number of times of the goal achievement success and the corresponding number of times of the goal achievement failure of the determined at least one conversation pattern.

Description

A method for determining an optimal conversation pattern for achieving a goal at a particular point in a conversation session associated with an AI service system, a method for determining a goal fulfillment prediction probability, and a computer-

The present disclosure relates to the determination of an interactive response by a conversation understanding AI service system, and more particularly to an interactive progress of a conversation that occurs with customers on a conversation understanding AI service system of a specific domain having a predetermined target task And how to determine the optimal dialog response for achieving the goal at each point in time.

Description of the Related Art [0002] In recent years, development of a conversation understanding AI service system capable of receiving and understanding natural language input of voice and / or text form by a user in accordance with technological development in artificial intelligence field, And utilization is gradually increasing. Meanwhile, in recent years, most companies and organizations operate a customer response center, and a dialogue understanding AI service system is being built for such a customer response center. These customer response centers often have a corresponding set of targeted tasks (e.g., a customer purchases the product or service for a given product and / or service center). The rate at which a customer accesses the Customer Response Center and reaches the target action of the Customer Response Center (the probability of achieving the goal) is the so-called conversion rate. The operator of the Customer Response Center, of course, I hope and explore various ways to do this. An agent in the Customer Response Center (a human agent or conversation AI Service System Agent) will interact with the customer who has accessed the Customer Response Center and convince the customer to reach the target action of the Customer Response Center.

Therefore, it is necessary to analyze conversation records between the customer and agent (person agent or conversation AI service system agent) to find a conversation pattern that can increase the conversion rate of the customer response center.

According to one aspect of the present disclosure, there is provided a method, comprising: on a conversation understanding AI service server configured to process a natural language conversation for each of a plurality of user terminals, a conversation session between a user terminal of one of a plurality of user terminals and a conversation understanding AI service server A method for determining a target achievement prediction probability for achieving a predetermined goal is provided. The method of the present disclosure includes receiving natural language input from a user terminal at a particular point in time; Determining an intent corresponding to the received natural language input; The conversation history at a specific point-of-conversation history represents the flow of intents between the user terminal and the conversation comprehension AI service server that occurred up to a certain point on the conversation session, and includes the determined intent at the end of the flow of intents. Determining; The template-template prepared in advance for the conversation history and the conversation comprehension AI service server includes a plurality of conversation patterns representing flows of the corresponding intents, and has a goal achievement frequency and a target failure frequency for each conversation pattern - determining, based on the at least one dialog pattern, one or more dialog patterns comprising a conversation history; And determining a target achievement prediction probability at a specific time point based on each of the corresponding goal achievement times and the target achievement failure times of the determined one or more conversation patterns.

According to one embodiment of the present disclosure, the target achievement prediction probability at a specific point in time is a ratio of the total sum of successes of achievement of the target achievement to the total sum of the corresponding goal achievement times and the target achievement failure times of the determined one or more conversation patterns, .

According to one embodiment of the present disclosure, the template may have a success rate determined based on the number of successes of the target achievement and the number of failure of the goal achievement of each dialog pattern for each of the plurality of conversation patterns.

In accordance with another aspect of the present disclosure there is provided a method, comprising: on a conversation understanding AI service server configured to process a natural language conversation for each of a plurality of user terminals, a conversation session between a user terminal of one of a plurality of user terminals and a conversation understanding AI service server There is provided a method for determining an optimal conversation pattern for achieving a predetermined goal at a time point. The method of the present disclosure includes receiving natural language input from a user terminal at a particular point in time; Determining an intent corresponding to the received natural language input; The conversation history at a specific point-of-conversation history represents the flow of intents between the user terminal and the conversation comprehension AI service server that occurred up to a certain point on the conversation session, and includes the determined intent at the end of the flow of intents. Determining; The template-template prepared in advance for the conversation history and the conversation comprehension AI service server includes a plurality of conversation patterns representing flows of the corresponding intents, and has a goal achievement frequency and a target failure frequency for each conversation pattern - determining, based on the at least one dialog pattern, one or more dialog patterns comprising a conversation history; Determining, for each of the determined one or more conversation patterns, a goal achievement probability determined based on the number of successes of the goal achievement and the number of failures of the goal achievement of each conversation pattern; And determining a conversation pattern having the highest goal achievement probability among the determined one or more conversation patterns as an optimal conversation pattern.

According to one embodiment of the present disclosure, the method may further comprise selecting a next conversation response on the conversation session from the determined optimal conversation pattern. According to one embodiment of the present disclosure, the conversation understanding AI service server includes a predetermined display device, and may be presented on the optimal conversation pattern determined through the display device or on the optimal conversation pattern, followed by the conversation response.

According to one embodiment of the present disclosure, a target may be associated with at least one of a predetermined product and / or service sale, a subscription, and a subscription.

According to one embodiment of the present disclosure, there is provided a method comprising collecting a plurality of conversation records, each conversation record comprising a series of correlated natural language inputs and responses associated with a target, occurring over a communication session, - the result of the achievement of; For each conversation record, determining each intent corresponding to each of the natural language inputs and responses of the conversation record and generating a corresponding set of flows of intents; Each conversation record corresponding to a conversation pattern of one of the predetermined conversation patterns in accordance with a flow of intents generated corresponding to the conversation record; Having a probability of achieving a goal corresponding to a conversation pattern, according to the results of the achievement; And generating a conversation template based on a result of associating the plurality of conversation records with one of the conversation patterns.

According to another aspect of the present disclosure there is provided a computer readable medium having stored thereon one or more instructions for causing a computer to perform any one of the methods described above when executed by the computer, A possible recording medium is provided.

According to the embodiment of the present disclosure, it is possible to analyze the conversation history recorded between the customer and the agent, and to determine various conversation patterns that may occur in the customer response center and the probability of achieving the goal for each conversation pattern. According to the embodiment of the present disclosure, it is possible to present the target achievement prediction probability at each point in the progress of the conversation progress between the customer and the agent, and to determine the optimal conversation pattern thereafter in order to improve the target achievement probability. Thus, according to the embodiments of the present disclosure, ultimately, the conversion rate of the customer response center can be improved.

1 is a schematic representation of a customer response center system environment 100 that may be implemented in accordance with one embodiment of the present disclosure.

FIG. 2 is a functional block diagram that schematically illustrates the functional configuration of the customer user terminal 102 of FIG. 1, in accordance with one embodiment of the present disclosure.

FIG. 3 is a functional block diagram that schematically illustrates the functional configuration of the AI service server 106 of FIG. 1 in accordance with one embodiment of the present disclosure.

FIG. 4 is a functional block diagram schematically illustrating the functional configuration of the dialog / task processing unit 304 of FIG. 3, according to one embodiment of the present disclosure.

FIG. 5 is an exemplary operational flow diagram performed by the dialog template creation / storage 308 of FIG. 3, in accordance with one embodiment of the present disclosure.

6 is a diagram conceptually showing an example of a template 600 for the conversation understanding AI service server 106. As shown in FIG.

FIG. 7 is an exemplary operational flow diagram performed by the dialog / task processing unit 304 of FIG. 3, in accordance with one embodiment of the present disclosure.

Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. Hereinafter, when it is judged that there is a possibility that the gist of the present disclosure may be unnecessarily blurred, a detailed description of known functions and configurations will be omitted. It is also to be understood that the following description is only an example of the present disclosure, and the present disclosure is not limited thereto.

The terminology used in this disclosure is used only to describe a specific embodiment and is not used to limit the present disclosure. For example, an element expressed in singular < Desc / Clms Page number 5 > terms should be understood as including a plurality of elements unless the context clearly dictates a singular value. It is to be understood that the term "and / or" as used in this disclosure encompasses any and all possible combinations of one or more of the listed items. It should be understood that the terms " comprises " or " having ", etc. used in the present disclosure are intended to specify that there exist features, numbers, steps, operations, elements, It is not intended to exclude the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof, by use.

In the embodiments of the present disclosure, 'module' or 'sub-module' means a functional part that performs at least one function or operation, and may be implemented in hardware or software, or a combination of hardware and software. Also, a plurality of "modules" or "sub-modules" may be integrated into at least one software module and implemented by at least one processor, except for "module" or "sub-module" have.

In the embodiment of the present disclosure, a 'conversation understanding AI service system' is a system in which a natural language input (for example, a command from a user in a natural language, a statement from a user in a natural language) input from a user through an interactive interaction via a natural- , Requests, questions, etc.) to determine the intent of the user and to provide the necessary actions, such as appropriate dialog response and / or predetermined task performance, based on the learned intent of the user The present invention is not limited to any particular type of information processing system.

In an embodiment of the present disclosure, the conversation response provided by the " Conversation Comprehension AI Service System " may be in the form of a visual, auditory and / or tactile (e.g., voice, sound, text, video, image, symbol, emoticon, hyperlink, Animation, various notices, motion, haptic feedback, and the like), and the like. In the embodiment of the present disclosure, the tasks performed by the " Dialogue AI Service System " may include, for example, searching for and providing information, progressing payment, membership, or any other type of task, ). &Lt; / RTI >

In accordance with an embodiment of the present disclosure, a 'conversation template' may be a template that includes all types of conversation patterns (flows of intents) between a customer and an agent that may occur on a customer response center system. In accordance with an embodiment of the present disclosure, a 'conversation template' is defined for each of the above-described conversation patterns as a goal of the customer response center system, that is, a target response center system, And / or the sale, subscription, and subscription of services, but are not limited thereby, and may be a variety of goals that have a particular purpose and may be explicitly identified as achievable) .

In addition, all terms used in the present disclosure, including technical or scientific terms, unless otherwise defined, have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. It should be understood that commonly used predefined terms are to be interpreted as having a meaning consistent with the contextual meaning of the related art and are not to be interpreted excessively or extensively unless explicitly defined otherwise in this disclosure .

Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings.

1 is a schematic representation of a customer response center system environment 100 that may be implemented in accordance with one embodiment of the present disclosure. As shown, the system environment 100 includes a plurality of customer user terminals 102, a communication network 104, and a conversation understanding AI service server 106.

According to one embodiment of the present disclosure, each of the plurality of customer user terminals 102 may be any user electronic device having wired or wireless communication capability. Each of the customer user terminals 102 may be a variety of wired or wireless communication terminals including, for example, a smart phone, a tablet PC, a music player, a smart speaker, a desktop, a laptop, a PDA, a game console, a digital TV, a set- But not limited to, the < / RTI > According to one embodiment of the present disclosure, each of the customer user terminals 102 can communicate with the AI service server 106 via the communication network 104, that is, send and receive necessary information. According to one embodiment of the present disclosure, each of the customer user terminals 102 may receive customer user input in the form of voice and / or text from the outside, and may communicate with the AI service server 106 via the communication network 104. [ (E.g., providing a specific conversation response and / or performing a specific task) corresponding to the above customer user input obtained through communication with the customer user terminal 102 (and / or processing in the customer user terminal 102) .

According to one embodiment of the present disclosure, the communication network 104 may comprise any wired or wireless communication network, e.g., a TCP / IP communication network. According to one embodiment of the present disclosure, the communication network 104 may include, for example, a Wi-Fi network, a LAN network, a WAN network, an Internet network, and the like, and the present disclosure is not limited thereto. In accordance with one embodiment of the present disclosure, the communication network 104 may be any of a variety of wired or wireless, including, but not limited to, Ethernet, GSM, EDGE, CDMA, TDMA, OFDM, Bluetooth, VoIP, Wi- May be implemented using a communication protocol.

According to one embodiment of the present disclosure, the conversation comprehension AI service server 106 may communicate with the customer user terminal 102 via the communication network 104. According to one embodiment of the present disclosure, the conversation understanding AI service server 106 receives customer user natural language input in the form of speech and / or text, for example, from the customer user terminal 102 via the communication network 104, The received natural language input may be processed based on the knowledge base model to determine the intent of the customer user.

According to one embodiment of the present disclosure, the conversation understanding AI service server 106 may have a predetermined goal. For example, according to one embodiment of the present disclosure, the conversation comprehension AI service server 106 may be a consultation center for the sale of certain products and / or services, It can be said that the target is achieved when the purchaser 102 purchases the corresponding product and / or service. According to another embodiment of the present disclosure, the conversation comprehension AI service server 106 includes a respective customer user terminal 102 that accesses the conversation comprehension AI service server 106 as a consultation center for inducing subscription or subscription, If you join or subscribe to this membership, you can assume that your goal is achieved.

According to one embodiment of the present disclosure, the conversation understanding AI service server 106 may communicate with one or more agent terminals (not shown), not shown, in a wired or wireless manner. For example, according to one embodiment of the present disclosure, voice and / or textual customer user natural language input from the customer user terminal 102 received on the conversation understanding AI service server 106 may be delivered to the agent terminal. According to one embodiment of the present disclosure, the conversation comprehension AI service server 106 may receive from the agent (e.g., a human consultant) via the agent terminal a natural language response of voice and / or text in response to the aforementioned customer user natural language input And may transmit the received natural language response to the customer user terminal 102 via the communication network 104. [

According to another embodiment of the present disclosure, the conversation understanding AI service server 106 may generate operation results in accordance with the user intent and communicate it to the customer user terminal 102 without communication with the agent terminal. According to one embodiment of the present disclosure, the conversation understanding AI service server 106 can perform an operation corresponding to the determined user intent based on a prepared conversation flow management model. According to one embodiment of the present disclosure, each action performed by the AI service server 106 may be an interaction response and / or task performance, e.g., corresponding to an intent of each user.

According to one embodiment of the present disclosure, the conversation understanding AI service server 106 may accumulate a plurality of conversation records sent and received between each of the customer user terminals 102 and the conversation understanding AI service server 106. In accordance with one embodiment of the present disclosure, each conversation record is associated with a target of the conversation comprehension AI service server 106 on one communication session established between the customer user terminal 102 and the conversation comprehension AI service server 106 (E.g., natural language inputs from the customer user terminals 102a-10n), responses from the agent terminal, or system responses by the conversation comprehension AI service server 106 Lt; / RTI > According to one embodiment of the present disclosure, the conversation comprehension AI service server 106 determines whether the conversation comprehension AI service server 106 has the same / similar domain as the AI service server 106, Multiple conversation records can be accumulated on other systems.

In accordance with one embodiment of the present disclosure, the conversation comprehension AI service server 106 receives each input / response (e.g., a customer user terminal 102a-10n ), A response from the agent terminal, or a system response by the conversation understanding AI service server 106). According to one embodiment of the present disclosure, the conversation comprehension AI service server 106 classifies each input / response into one of a predetermined set of intents via keyword analysis, as described above, Lt; RTI ID = 0.0 > of intents. &Lt; / RTI >

As described above, the conversation comprehension AI service server 106 may have a predetermined goal, and each conversation record, that is, a customer user input from an associated customer user terminal 102a-10n that is transmitted and received on one communication session And a corresponding series of responses of the agent's terminal or system responses by the conversation comprehension AI service server 106 may each have a result of whether or not the goal of the conversation comprehension AI service server 106 has been achieved . According to one embodiment of the present disclosure, the conversation comprehension AI service server 106 may generate a plurality of conversation conversations with a predetermined number of conversation patterns (each conversation pattern is a predetermined pattern representing the flow of intents, ), Respectively. According to one embodiment of the present disclosure, the conversation comprehension AI service server 106 may record the number of times the goal was achieved and the number of times the goal was not achieved for each conversation pattern, . According to one embodiment of the present disclosure, the conversation understanding AI service server 106 associates the above-described conversation patterns with a goal achievement probability (or success and failure times) for each conversation pattern, And a template of conversation patterns for the conversation comprehension AI service server 106, including both the goal achievement probabilities. According to another embodiment of the present disclosure, the conversation comprehension AI service server 106 may include a conversation comprehension AI service < RTI ID = 0.0 > And generate and store a template of dialog patterns for the server 106.

According to one embodiment of the present disclosure, one of the customer user terminals 102 accesses the conversation comprehension AI service server 106 to establish a communication session over the communication network 104, It is possible to determine an optimum conversation pattern after the specific point in time to improve the target achievement probability at each point in the course of conversation. According to one embodiment of the present disclosure, the conversation understanding AI service server 106 can also determine the optimal conversation pattern at each point in the conversation progression as described above, and at the same time, the conversation comprehension AI service server 106 ) Can be determined.

Specifically, the conversation comprehension AI service server 106 determines whether or not a specific input at a specific point in time (e.g., at a point in time when a specific input has occurred in an ongoing conversation, , A series of intents that have occurred between the customer user terminal 102 and the conversation AI service server 106 until reaching that point in time (intents corresponding to the input / responses) ), And based on the template of the conversation patterns generated in the above, it is possible to grasp the conversation patterns that can be developed in the future, that is, conversation patterns that can be generated in the future. The conversation understanding AI service server 106 recognizes the conversation patterns that can be generated in the future and then determines the target achieving probability (the success rate determined by the number of successes of the target achievement and the number of failures of the target achievement of each conversation pattern) . According to one embodiment of the present disclosure, the conversation understanding AI service server 106 compares the target achievement probabilities of the possible conversation patterns after the specific point in time, As shown in FIG. According to one embodiment of the present disclosure, the conversation comprehension AI service server 106 may also determine, based on the number of successes and failures of the goal achievement for all of the future possible conversation patterns, , It is possible to determine the target achievement prediction probability at the specific time point.

FIG. 2 is a functional block diagram that schematically illustrates the functional configuration of the customer user terminal 102 shown in FIG. 1, according to one embodiment of the present disclosure. The customer user terminal 102 includes a user input receiving module 202, a sensor module 204, a program memory module 206, a processing module 208, a communication module 210, (212).

According to one embodiment of the present disclosure, the user input receiving module 202 is configured to receive various types of input from a user, such as natural language input (such as voice input and / or text input, and additionally other types of input Can be received. According to one embodiment of the present disclosure, the user input receiving module 202 includes, for example, a microphone and an audio circuit, and can acquire a user voice input signal through a microphone and convert the obtained signal into audio data. According to one embodiment of the present disclosure, the user input receiving module 202 may include various types of input devices such as various pointing devices such as a mouse, joystick, trackball, keyboard, touch panel, touch screen, stylus, , And can acquire a text input and / or a touch input signal inputted from a user through these input devices. According to one embodiment of the present disclosure, the user input received at the user input receiving module 202 may be associated with performing certain tasks, such as performing certain applications or retrieving certain information, It is not. According to another embodiment of the present disclosure, the user input received at the user input receiving module 202 may require only a simple conversation response, regardless of any application execution or retrieval of information. According to another embodiment of the present disclosure, the user input received at the user input receiving module 202 may relate to a simple statement for unilateral communication.

According to one embodiment of the present disclosure, the sensor module 204 comprises one or more different types of sensors, and through these sensors, status information of the customer user terminal 102, such as the physical Status, software and / or hardware status, or information regarding the environmental conditions of the customer user terminal 102, and the like. According to one embodiment of the present disclosure, the sensor module 204 may include an optical sensor, for example, and may sense the ambient light condition of the customer user terminal 102 via the optical sensor. According to one embodiment of the present disclosure, the sensor module 204 may include, for example, a movement sensor and may sense movement of the corresponding customer user terminal 102 via the movement sensor. According to one embodiment of the present disclosure, the sensor module 204 includes, for example, a velocity sensor and a GPS sensor, and through these sensors, the position and / or orientation of the corresponding customer user terminal 102 may be sensed. It should be noted that, according to another embodiment of the present disclosure, the sensor module 204 may include other various types of sensors, including temperature sensors, image sensors, pressure sensors, touch sensors, and the like.

According to one embodiment of the present disclosure, the program memory module 206 may be any storage medium that stores various programs that may be executed on the customer user terminal 102, such as various application programs and associated data, and the like. In accordance with one embodiment of the present disclosure, program memory module 206 may include one or more applications, such as a telephone dialer application, an email application, an instant messaging application, a camera application, a music playback application, a video playback application, an image management application, , And data related to the execution of these programs. According to one embodiment of the present disclosure, program memory module 206 may be configured to include volatile or nonvolatile memory of various types such as DRAM, SRAM, DDR RAM, ROM, magnetic disk, optical disk, flash memory, .

According to one embodiment of the present disclosure, the processing module 208 may communicate with each component module of the customer user terminal 102 and perform various operations on the customer user terminal 102. According to one embodiment of the present disclosure, the processing module 208 can drive and execute various application programs on the program memory module 206. [ According to one embodiment of the present disclosure, the processing module 208 may receive signals obtained from the user input receiving module 202 and the sensor module 204, if necessary, and perform appropriate processing on these signals have. According to one embodiment of the present disclosure, the processing module 208 may, if necessary, perform appropriate processing on signals received from the outside via the communication module 210.

According to one embodiment of the present disclosure, the communication module 210 enables the customer user terminal 102 to communicate with the conversation understanding AI service server 106 via the communication network 104 of FIG. According to one embodiment of the present disclosure, the communication module 210 may be configured to communicate with a user terminal such as, for example, a user input receiving module 202 and a sensor module 204 via a communication network 104, To be transmitted to the server 106. In accordance with one embodiment of the present disclosure, the communication module 210 may provide a response including a natural language response in the form of various signals, e.g., voice and / or text, received from the conversation understanding AI service server 106 via, for example, Signals, various control signals, and the like, and can perform appropriate processing according to a predetermined protocol.

According to one embodiment of the present disclosure, the response output module 212 may output a response corresponding to a user input in various forms, such as time, audible and / or tactile. According to one embodiment of the present disclosure, the response output module 212 includes various display devices, such as a touch screen based on technology such as LCD, LED, OLED, QLED, etc., Such as text, symbols, video, images, hyperlinks, animations, various notices, etc., to the user. According to one embodiment of the present disclosure, the response output module 212 includes a speaker or a headset, for example, and provides an audible response, e.g., voice and / or acoustic response, can do. According to one embodiment of the present disclosure, the response output module 212 includes a motion / haptic feedback generator, through which a tactile response, e.g., motion / haptic feedback, can be provided to the user. It should be appreciated that, in accordance with one embodiment of the present disclosure, the response output module 212 may concurrently provide any combination of two or more of a text response, a voice response and a motion / haptic feedback corresponding to a user input.

FIG. 3 is a functional block diagram that schematically illustrates the functional configuration of the AI service server 106 of FIG. 1 in accordance with one embodiment of the present disclosure. The conversation understanding AI service server 106 includes a communication module 302, a dialog / task processing section 304, a conversation record accumulation section 306, and a dialog template creation / storage section 308 .

According to one embodiment of the present disclosure, the communication module 302 is configured to communicate with the AI service server 106 via the communication network 104, in accordance with any wired or wireless communication protocol, To communicate with a terminal (not shown). According to one embodiment of the present disclosure, the communication module 302 is capable of receiving, via the communication network 104, voice input and / or text input, etc., received from the customer user terminal 102, have. According to one embodiment of the present disclosure, the communication module 302 may communicate with the customer user terminal 102 via the communication network 104 with or without voice input and / or text input from the user, The status information of the customer user terminal 102 transmitted from the customer user terminal 102 can be received. According to one embodiment of the present disclosure, the state information may include various state information (e.g., the physical state of the customer user terminal 102) associated with the customer user terminal 102 at the time of speech input from the user and / Software and / or hardware status of the customer user terminal 102, environmental status information around the customer user terminal 102, etc.). In accordance with one embodiment of the present disclosure, the communication module 302 may also include an interaction response (e.g., a natural-language interaction response in the form of voice and / or text) generated in response to the received customer user input and / May perform the appropriate actions necessary to communicate the signal, via the communication network 104, to the customer user terminal 102.

According to one embodiment of the present disclosure, the dialog / task processing unit 304 receives user natural language input from the customer user terminal 102 via the communication module 302, and based on predetermined knowledge base models prepared in advance The intent of the user corresponding to the user natural language input can be determined by processing this. According to one embodiment of the present disclosure, the dialog / task processing unit 304 may also provide an action consistent with the determined user ' s tent, e.g., appropriate dialog response and / or task performance.

According to one embodiment of the present disclosure, when the tent of the user from the customer user is determined during the conversation progress, the conversation / task processing unit 304 acquires the conversation history up to now and the conversation understanding AI service server 106 By referring to the template, it is possible to grasp the dialog patterns that can be generated in the future, and to determine the target achievement probability in each of the dialog patterns. According to one embodiment of the present disclosure, the dialog / task processing unit 304 compares the determined target achievement probabilities for each of the possible dialog patterns after the specific point in time, Can be determined by the dialog pattern of FIG. According to one embodiment of the present disclosure, the conversation / task processing unit 304 also calculates a target achievement prediction probability at the corresponding point in time based on the number of times of success and the number of failures of the goal achievement with respect to all of the conversation patterns that can be generated in the future Can be determined. . According to one embodiment of the present disclosure, the dialog / task processing unit 304 can provide a response according to the determined optimal dialog pattern as a response corresponding to the above-described user natural language input, for example, as a natural language in voice or text form have. According to another embodiment of the present disclosure, the dialog / task processing unit 304 provides a response to the consultant terminal (not shown) according to the determined optimal dialog pattern as a response corresponding to the user natural language input described above, Counselors on the counselor's terminal can refer to them.

According to one embodiment of the present disclosure, the conversation record accumulation unit 306 includes a conversation comprehension AI service server 106 and a customer user terminal 102, which are obtained on the conversation comprehension AI service server 106 in Fig. 3 (E.g., each conversation record is made up of inputs from customer user terminals 102a-10n and responses from an agent terminal or system responses by conversation comprehension AI service server 106) And may include a series of conversation flow records). In accordance with one embodiment of the present disclosure, the conversation record accumulation unit 106 may be configured to perform the conversation with the AI service server 106, which is not generated by the AI service server 106, Lt; RTI ID = 0.0 > similar < / RTI > domain.

According to one embodiment of the present disclosure, the conversation template creation / storage unit 308 analyzes each conversation record on the conversation record accumulation unit 306 and stores each input / response (e.g., (E.g., input from an agent terminal 102a-10n, response from an agent terminal, or system response by a conversation understanding AI service server 106) to one of predetermined predetermined intents. According to one embodiment of the present disclosure, the conversation template creation / storage unit 308 performs keyword analysis on each input / response of each conversation record on the conversation record storage unit 306, for example, May be classified into one of the predetermined intents. According to one embodiment of the present disclosure, the conversation understanding AI service server 106 determines the intent of each input / response on the conversation record for each conversation record on the conversation record storage unit 306, as described above And thereby change to a sequence of predetermined intents corresponding to the respective conversation record. Thus, according to one embodiment of the present disclosure, the conversation template creation / storage unit 308 stores the conversation records accumulated on the conversation record storage unit 306 in a predetermined number of dialog patterns (predetermined Pattern), respectively.

As discussed above, each conversation record may have its own outcome as to the predetermined goal of the conversation comprehension AI service server 106, i. E. As a result of the achievement success or failure. Therefore, the dialog template generating / storing unit 308 can record the number of times the target is achieved and the number of times that the target is not achieved for each conversation pattern, and acquires and records the probability that the target task of the conversation pattern is achieved arithmetically can do. According to one embodiment of the present disclosure, the conversation template creation / storage unit 308 associates each of the generated conversation patterns with the goal achievement probability (or success and failure times) for each conversation pattern, And generate and store a template for the conversation comprehension AI service server 106, which includes both conversation patterns and goal fulfillment probabilities. According to another embodiment of the present disclosure, the conversation template creation / storage unit 308 generates a conversation template having a plurality of conversation patterns, each of which includes conversation patterns whose occurrence count exceeds a predetermined reference value, And generate and store a template of dialog patterns for the service server 106. [

Although not specifically shown, according to one embodiment of the present disclosure, the dialog template generating / storing unit 308 may include a predetermined display device, visualizes a template of the generated dialog patterns, , The disclosure of which is not so limited.

FIG. 4 is a functional block diagram schematically illustrating the functional configuration of the dialog / task processing unit 304 of FIG. 3, according to one embodiment of the present disclosure. Task processing unit 304 includes a Speech-To-Text (STT) module 402, a Natural Language Understanding (NLU) module 404, a user database 406 A dialogue understanding knowledge base 408, a dialogue management module 410, a dialogue generation module 412, and a speech-to-speech (TTS) module 414.

In accordance with one embodiment of the present disclosure, the STT module 402 is capable of receiving speech input during user input received via communication module 302 and converting the received speech input into text data based on pattern matching, have. According to one embodiment of the present disclosure, the STT module 402 may extract a feature from a user's speech input to generate a feature vector sequence. According to one embodiment of the present disclosure, the STT module 402 may be implemented using a DTW (Dynamic Time Warping) method, an HMM model (Hidden Markov Model), a GMM model (Gaussian-Mixture Mode), a deep neural network model, For example, a sequence of words, based on various statistical models of the speech recognition results. According to one embodiment of the present disclosure, the STT module 402 may refer to each user characteristic data of the user database 406, described below, when converting the received voice input into text data based on pattern matching .

According to one embodiment of the present disclosure, the NLU module 404 may receive text input from the communication module 302 or the STT module 402. According to one embodiment of the present disclosure, the textual input received at the NLU module 404 may be transmitted to the user user terminal 102 via the user text input or communication module (e.g., 302 may be a text recognition result, e.g., a sequence of words, generated by the STT module 402 from the user speech input received at. According to one embodiment of the present disclosure, the NLU module 404 may be configured to receive status information associated with the user input, such as with or after receipt of the text input, such as the status of the customer user terminal 102 Information and the like can be received. As described above, the status information may include various status information (e.g., the physical (physical) information of the customer user terminal 102) related to the customer user terminal 102 at the time of user input and / State of the software, and / or hardware status, environmental condition information around the customer user terminal 102, etc.).

In accordance with one embodiment of the present disclosure, the NLU module 404 may map the received text input to one or more user-defined intents based on the dialog understanding knowledge base 408. Where the user intent may be associated with a series of operations (s) that can be understood and performed by the AI service server 106 of the conversation understanding according to the user's tent. According to one embodiment of the present disclosure, the NLU module 404 may refer to the status information described above in mapping the received textual input to one or more user intents. According to one embodiment of the present disclosure, the NLU module 404 may refer to each user characteristic data of the user database 406, described below, in mapping the received text input to one or more user intents.

According to one embodiment of the present disclosure, the user database 406 may be a database that stores and manages characteristic data for each user. According to one embodiment of the present disclosure, the user database 406 may include, for example, previous conversation history of the user for each user, pronunciation feature information of the user, user lexical preference, location of the user, And may include various user-specific information.

According to one embodiment of the present disclosure, as described above, the STT module 402 may determine each user characteristic data of the user database 406, e.g., each user-specific pronunciation characteristic, , More accurate text data can be obtained. According to one embodiment of the present disclosure, the NLU module 404 may determine a more accurate user tent determination by referring to each user characteristic data of the user database 406, e.g., characteristics or contexts for each user, can do.

In the figure, a user database 406 for storing and managing characteristic data for each user is shown as being placed in the conversation understanding AI service server 106, but this disclosure is not limited thereto. According to another embodiment of the present disclosure, a user database that stores and manages characteristic data for each user may be present at, for example, the customer user terminal 102 and may include a customer user terminal 102 and a conversation comprehension AI service server 106. [ As shown in FIG.

According to one embodiment of the present disclosure, the conversation understanding knowledge base 408 may include, for example, a predefined ontology model. According to one embodiment of the present disclosure, an ontology model can be represented, for example, in a hierarchical structure between nodes, each node having an "intent" node corresponding to the user's intent or a &Quot; Attributes " node that is linked directly to an " Attributes "node or a " Attributes" node of an "Attributes" According to one embodiment of the present disclosure, the " intent "node and the" attribute "nodes directly or indirectly linked to the" intent "node can constitute one domain and the ontology comprises a set of such domains . According to one embodiment of the present disclosure, the conversation understanding knowledge base 308 includes, for example, domains corresponding to all intents that can be understood by the conversation understanding AI service server 106 and perform corresponding actions . It should be noted that, according to one embodiment of the present disclosure, the ontology model can be dynamically changed by addition or deletion of nodes, or modification of relations between nodes.

According to one embodiment of the present disclosure, the intent nodes and attribute nodes of each domain in the ontology model may be associated with words and / or phrases associated with the corresponding user's tent or attributes, respectively. According to one embodiment of the present disclosure, the conversation understanding knowledge base 408 includes an ontology model 408, which may include an ontology model, for example, in a lexical dictionary form (specifically, , And the NLU module 404 may determine the user intent based on the ontology model implemented in the lexical dictionary form. For example, according to one embodiment of the present disclosure, the NLU module 404, upon receipt of a textual input or sequence of words, can determine which nodes in a domain within each of the words in the sequence are associated with which nodes in the ontology model, Based on such a determination, it is possible to determine the corresponding domain, i. E. The user tent.

In accordance with one embodiment of the present disclosure, the conversation management module 410 may generate a corresponding series of operational flows in accordance with the user ' s tent determined by the NLU module 404. In accordance with one embodiment of the present disclosure, the conversation management module 410 may be configured to perform any action (e.g., based on the user's intent received from the NLU module 404) based on a predetermined conversation flow management model E.g., what dialog response and / or task execution should be performed, and generate a corresponding detailed action flow.

According to one embodiment of the present disclosure, when the user intent is determined, the dialog management module 410 refers to the previous conversation history and the template for the conversation understanding AI service server 106 described above, It is possible to grasp the conversation patterns, and to determine and provide the target achievement prediction probability at that point in time. According to one embodiment of the present disclosure, when the user intent is determined, the dialog management module 410 refers to the previous conversation history and the template for the conversation understanding AI service server 106 described above, It is possible to grasp the conversation patterns and to determine the target achieving probability in each of the conversation patterns. According to one embodiment of the present disclosure, the dialogue management module 410 compares the determined goal achievement probabilities for each of the possible dialog patterns that occur after the specific point in time, It is possible to determine the optimal dialog pattern in FIG. According to one embodiment of the present disclosure, the conversation management module 410 determines whether to perform an action flow (e.g., any conversation response and / or task performance) as to which action to perform based on the determined optimal conversation pattern A flow relating to whether or not to do so).

According to one embodiment of the present disclosure, the dialog generation module 412 may generate the necessary dialog response based on the operation flow generated by the dialogue management module 410. [ According to one embodiment of the present disclosure, the dialog generation module 412 is configured to generate user interaction data (e.g., user's previous conversation history, user's pronunciation feature information, Lexical preference, user's location, set language, contact / friend list, previous user conversation history for each user, etc.).

In accordance with one embodiment of the present disclosure, the TTS module 414 may receive an interactive response that is generated by the dialog generation module 412 to be transmitted to the customer user terminal 102. [ The conversation response received at the TTS module 414 may be a natural word or a sequence of words having a textual form. According to one embodiment of the present disclosure, the TTS module 414 may convert the input of the received textual form into speech form, according to various types of algorithms.

In the embodiments of the present disclosure described above with reference to Figs. 1-4, a particular module has been described as if it were performing certain operations for convenience, but the present disclosure is not so limited. It should be noted that, in accordance with another embodiment of the present disclosure, the operations described above as performed by any particular module in the above description can be performed by separate and distinct modules, respectively.

5 is an exemplary operational flow diagram performed by the dialog template creation / storage 306 of FIG. 3, in accordance with one embodiment of the present disclosure.

In step 502, the conversation template creation / storage unit 308 creates a conversation template for each of the conversation records collected in any of various ways (specifically, for each entry of each conversation record, e.g., for each sentence or phrase of each intent unit) Records) can be analyzed. According to one embodiment of the present disclosure, the dialog template generating / storing unit 308 analyzes each input record on each conversation record according to a predetermined criterion and classifies it into one of predetermined intent groups. The above input may be obtained, for example, from each conversation record sent and received between the conversation comprehension AI service server 106 and the customer user terminal 102, respectively, According to one embodiment of the present disclosure, the above input may be obtained from conversation records collected by any other method.

Then, at step 504, the dialog template creation / storage unit 308 creates, based on the analysis result at step 502, a series of predetermined intents corresponding to each conversation record, for example, It is possible to generate a predetermined dialog pattern. Next, at step 506, the conversation template creation / storage unit 308 groups the conversation records corresponding to the same conversation pattern, and for each conversation record belonging to the group in the same group, It is possible to determine the number of the goal achievement success conversation records and the number of the goal attainment failure conversation records depending on the result about the goal of the server 106, that is, success or failure.

In step 508, the conversation template creation / storage unit 308 associates the number of goal achievement / failure conversation records (or the goal achievement probability) for each of the generated conversation patterns with each other, And a number of goals achievement / failure conversation records (or goal achievement probabilities) associated with each conversation pattern. In step 510, the conversation template creation / storage unit 308 visualizes the number of conversation patterns stored in step 508 and the number of success / failure conversation records for each conversation pattern, and displays And this disclosure is not so limited.

As shown, the patterns of conversation patterns, that is, the sequence of intent flows that can be performed on the conversation comprehension AI service server 106, included in the template 600, can be classified into nine categories. Each of the rows of the dialog pattern column represents the intent of the intent, and each row of the dialog pattern column represents the intent of the intent, as shown in the rows of the dialog pattern column. As shown, each of A, B, C, D, E, F, G, H, And a sequential flow between them. As shown, each conversation pattern is also associated with a number of successes and failures, and a corresponding success rate. For example, in the case of the first row, the first row has a pattern of A-> B-> C-> D- > E among the conversation patterns, , So the success rate is indicated as 0.7. In the case of the last ninth line conversation pattern, that is, A-> D-> C-> K-> G, the success rate is 1 and the failure rate is 0 and the success rate is 1. According to one embodiment of the present disclosure, for example, the template can be configured to include only the dialog patterns whose occurrence count exceeds a predetermined reference value (for example, 30), and in this case, A-> D-> It should be noted that the pattern C-> K-> G can be excluded from the template.

It should be understood that the matter shown in FIG. 6 is merely a conceptual illustration of an extremely simplified dialog pattern to aid understanding of the embodiments of the present disclosure, and is not intended to limit the present disclosure. It should be noted that according to another embodiment of the present disclosure, various types of dialog patterns may appear in various forms.

According to one embodiment of the present disclosure, the dialog / task processing unit 304 may, at step 702, receive a user natural language input from the customer user terminal 102 of FIG. Then, at step 704, the user natural language input received above may be processed based on predetermined knowledge base models previously prepared to determine the intent of the user corresponding to the user natural language input.

Then, at step 706, the dialog / task processing unit 304 refers to the previous conversation history and a template (for example, the template 600 of FIG. 6) for the conversation understanding AI service server 106 So that future dialog patterns that can be generated can be obtained. According to one embodiment of the present disclosure, for example, if the intent determined at step 704 is B and the previous conversation history is a pattern of A- > B, then at step 706, As possible dialog patterns, 1, 2, 5, and 6 rows, respectively, can be obtained.

Then, at step 708, the dialog / task processing unit 304 may determine the target achievement probability of each of the dialog patterns obtained in step 706 (i.e., future dialog patterns). For example, the target achievement probability of each conversation pattern may be a success rate determined based on the number of successes of goal achievement and the number of failure of goal achievement with respect to the conversation pattern. Then, the procedure proceeds to step 710, and the dialog / task processing unit 304 compares the determined target achievement probability with respect to each of the possible dialog patterns to be generated in the future, It is possible to determine the optimum dialog pattern at the time. According to one embodiment of the present disclosure, for example, the conversation / task processing unit 304 determines that the conversation history up to now is a pattern of A- > B, and that conversation patterns that can be generated in the future are 1, 2, 5, It is possible to determine that the two lines having the highest success rate among these are the optimal dialog pattern at the current point of time.

In step 712, in accordance with one embodiment of the present disclosure, the dialog / task processing unit 304 determines whether the goal attainment of the entire conversation patterns (i.e., future possible dialog patterns) The target achievement prediction probability at the time point can be determined based on the number of successes and the number of failures. According to one embodiment of the present disclosure, if 1, 2, 5, and 6 rows, respectively, were obtained as possible dialog patterns in the future, in step 706, the total number of successes in these rows was 242, Is 117 times, and the predicted probability of achieving the target at that point of time can be about 0.67.

As will be appreciated by those skilled in the art, the present disclosure is not limited to the examples described herein, but may be variously modified, rearranged, and replaced within the scope of the present disclosure. It should be understood that the various techniques described herein may be implemented in hardware or software, or a combination of hardware and software.

A computer program according to an embodiment of the present disclosure may be stored in a storage medium readable by a computer processor or the like, for example, a non-volatile memory such as an EPROM, EEPROM, or flash memory device, a magnetic disk such as an internal hard disk and a removable disk, CDROM disks, and the like. Also, the program code (s) may be implemented in assembly language or machine language. All such modifications and variations that fall within the true spirit and scope of this disclosure are intended to be embraced by the following claims.

Claims

On a conversation understanding AI service server configured to handle a natural language conversation for each of a plurality of user terminals, at a certain point during the conversation session between one user terminal of the plurality of user terminals and the conversation understanding AI service server, As a method for determining a target achievement prediction probability for achieving,

Receiving a natural language input from the user terminal at the specific time;

Determining an intent corresponding to the received natural language input;

A conversation history at the specific time point, the conversation history indicating a flow of intents between the user terminal and the conversation comprehension AI service server occurring up to the specific point on the conversation session, and at the end of the flow of intents Determining that the determined intent is included;

A template prepared in advance for the conversation history and the conversation understanding AI service server, the template including a plurality of conversation patterns representing the flow of corresponding intents, Determining one or more conversation patterns including the conversation history from among the conversation patterns based on the number of times of failure; And

Determining the target achievement prediction probability at the specific time point based on each of the corresponding goal achievement number of times and the target failure number of times of the determined one or more conversation patterns

And determining a target achievement prediction probability.
The method according to claim 1,

Wherein the target achievement prediction probability at the specific time point corresponds to a ratio of a total sum of the number of successes of the target achievement to the total sum of the corresponding goal achievement number of times and the target achievement failure number of the one or more determined conversation patterns How to determine the target achievement probability.
The method according to claim 1,

Wherein the template has a success rate determined based on the target achievement number of times and the target achievement failure number of each conversation pattern for each of the plurality of conversation patterns.
The method according to claim 1,

Wherein the target is associated with at least one of a predetermined product and / or service sale, a subscription, and a subscription.
A conversation comprehended to process a natural language conversation for each of a plurality of user terminals on a AI service server, a user terminal of one of the plurality of user terminals and a conversation comprehension AI service server, The method comprising:

Receiving a natural language input from the user terminal at the specific time;

Determining an intent corresponding to the received natural language input;

A conversation history at the specific time point, the conversation history indicating a flow of intents between the user terminal and the conversation comprehension AI service server occurring up to the specific point on the conversation session, and at the end of the flow of intents Determining that the determined intent is included;

A template prepared in advance for the conversation history and the conversation understanding AI service server, the template including a plurality of conversation patterns representing the flow of corresponding intents, Determining one or more conversation patterns including the conversation history from among the conversation patterns based on the number of times of failure;

Determining, for each of the determined one or more dialog patterns, a target achievement probability determined based on the number of successes of the target achievement and the number of failures of the target achievement of the respective dialogue patterns; And

Determining an optimal dialog pattern having the highest goal achievement probability among the determined one or more dialog patterns as an optimal dialog pattern,

And determining an optimal dialog pattern.
6. The method of claim 5,

And selecting a next conversation response on the conversation session from the determined optimal conversation pattern.
The method according to claim 6,

Wherein the conversation understanding AI service server includes a predetermined display device and presents the next dialog response selected on the determined optimal dialog pattern or the optimal dialog pattern through the display device .
The method according to claim 1,

Wherein the target is associated with at least one of a predetermined product and / or service sale, a subscription, and a subscription.
The method according to claim 1,

Collecting a plurality of conversation records, each of the conversation records comprising a series of correlated natural language inputs and responses associated with the target, occurring over a communication session, -;

Determining, for each of the conversation records, each intent corresponding to each of the natural language inputs and responses of the conversation record and generating a corresponding series of intent flows;

Associating each of the conversation records with one of the predetermined conversation patterns in accordance with the flow of the generated intents corresponding to the conversation record, wherein the conversation pattern includes a conversation record corresponding to the conversation pattern Having a probability of achieving the target corresponding to the conversation pattern, in accordance with the results of achieving the goal; And

Further comprising generating a dialog template based on a result of associating the plurality of conversation records with one of the conversation patterns.
Readable medium having stored thereon one or more instructions for causing a computer to perform the method of any one of claims 1 to 9 when executed by a computer, .