Dialogue control device, dialogue control system, dialogue control method, and dialogue control program

Abstract

To provide a technique for succeeding dialogue with AI to dialogue with a person so as to achieve seamless escalation.SOLUTION: A dialogue control device disclosed herein can be connected to each of a first user terminal, a plurality of second user terminals, and a conversational AI using a large language model and comprises: a dialogue control unit configured to control first dialogue between a first user using the first user terminal and the conversational AI and second dialogue between the first user and a second user using one of the second user terminals; and a determination unit for determining whether or not to execute a succession to the second dialogue, where the dialogue control unit, upon determination by the determination unit that the succession should be executed during the first dialogue, causes the conversational AI to select second user candidates from a list comprising a plurality of second users according to content of the first dialogue and lets the second dialogue start with a second user selected from among the candidates.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a dialogue control device, a dialogue control system, a dialogue control method, and a dialogue control program. [Background technology]

[0002] With the development of large-scale language models, chatbots are increasingly being used as interfaces in contact centers, etc. In interactive information systems, chatbots can be used to handle simple inquiries and guidance tasks, making it possible to efficiently allocate human resources.

[0003] Patent Document 1 describes an example of a technology that checks a customer database that stores customer identification information and importance in association with each other, and notifies a supervisor terminal if it is determined that the call recipient is an important customer. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2012-120120 Summary of the Invention [Problem to be solved by the invention]

[0005] The above system makes it easier to prepare for a sudden change of the person in charge, such as a supervisor, during a conversation with an important customer. On the other hand, the decision to change the conversation must be made depending on the content of the conversation, and when conducting an automated conversation using a chatbot, etc., it is necessary to appropriately switch the person in charge from the chatbot to an operator. Therefore, it is necessary to make it possible to hand over responsibility to a human (operator) based on the history of the conversation.

[0006] The present invention has been made in consideration of the above points, and aims to provide a technology that enables seamless escalation by handing over dialogue with AI to dialogue with humans. [Means for solving the problem]

[0007] The present application includes a number of means for solving at least part of the above problems, examples of which are as follows.

[0008] In order to solve the above problem, a dialogue control device according to one embodiment of the present invention is a dialogue control device that can be connected to each of a first user terminal, a plurality of second user terminals, and a dialogue AI using a large-scale language model, and is equipped with a dialogue control unit that controls a first dialogue between a first user using the first user terminal and the dialogue AI, and a second dialogue between the first user and a second user using one of the second user terminals, and a judgment unit that determines whether to execute handover to any of the second dialogues during the first dialogue, wherein the dialogue control unit, when the judgment unit determines to execute handover during the first dialogue, causes the dialogue AI, which has recorded the content of the first dialogue, to select one or more candidates for the second user to conduct the second dialogue from a list including a plurality of the second users, depending on the content of the first dialogue, and starts the second dialogue with one of the second users selected from the candidates.

[0009] In addition, in the above-mentioned dialogue control device, the list may be associated with a predetermined attribute for each second user, and the dialogue control unit may cause the dialogue AI to select candidates for the second user who have the attribute that is highly relevant to the summary of the first dialogue.

[0010] In addition, in the above-mentioned dialogue control device, the dialogue control unit may select the second user from the selected candidates for the second user using a model that has been trained using the performance of the second user as an explanatory variable and the satisfaction of the first user as a target variable.

[0011] In the above-described dialogue control device, the dialogue control unit may select the second user who is available to respond from among the selected second user candidates.

[0012] In addition, in the above-mentioned dialogue control device, the dialogue control unit may display the selected candidates for the second user on the first user terminal and accept a selection of the second user from the first user.

[0013] In addition, in the above-mentioned dialogue control device, the dialogue control unit may notify the second user terminal of the selected candidate for the second user and select the second user by accepting the candidacy.

[0014] In addition, a dialogue control system according to another aspect of the present invention is a dialogue control system including a dialogue control device connectable to each of a first user terminal, a plurality of second user terminals, and a dialogue AI using a large-scale language model, and is characterized in that the dialogue control system performs a dialogue control step of controlling a first dialogue between a first user using the first user terminal and the dialogue AI, and a second dialogue between the first user and a second user using any of the second user terminals, and a judgment step of determining whether to execute handover to any of the second dialogues during the first dialogue, and in the dialogue control step, if it is determined in the judgment step during the first dialogue that handover will be executed, the dialogue AI that recorded the content of the first dialogue selects one or more candidates for the second user to conduct the second dialogue from a list including a plurality of the second users, depending on the content of the first dialogue, and starts the second dialogue with one of the second users selected from the candidates.

[0015] Another aspect of the present invention provides a dialogue control method using a dialogue control device connectable to a first user terminal, a plurality of second user terminals, and a dialogue AI using a large-scale language model, wherein the dialogue control device is provided with a processing unit, and the processing unit performs a dialogue control step of controlling a first dialogue between a first user using the first user terminal and the dialogue AI, and a second dialogue between the first user and a second user using one of the second user terminals, and a judgment step of determining whether to execute handover to any of the second dialogues during the first dialogue, wherein the dialogue control step is characterized in that, when it is determined in the judgment step during the first dialogue that handover is to be executed, the dialogue AI, which has recorded the content of the first dialogue, selects one or more candidates for the second user to engage in the second dialogue from a list including a plurality of the second users, according to the content of the first dialogue, and starts the second dialogue with one of the second users selected from the candidates.

[0016] Another aspect of the present invention provides a dialogue control program that causes an information processing device to control a dialogue, the information processing device having a processing unit, and causing the processing unit to perform a dialogue control step of controlling a first dialogue between a first user using a first user terminal and a dialogue AI using a large-scale language model, and a second dialogue between the first user and a second user using one of a plurality of second user terminals, and a judgment step of determining whether to execute handover to any of the second dialogues during the first dialogue, wherein the dialogue control step is characterized in that, when it is determined in the judgment step during the first dialogue that handover is to be executed, the dialogue AI that recorded the content of the first dialogue selects one or more candidates for the second user to engage in the second dialogue from a list including a plurality of the second users, depending on the content of the first dialogue, and starts the second dialogue with one of the second users selected from the candidates. [Effects of the Invention]

[0017] According to the present invention, it is possible to provide a technology that enables seamless escalation by handing over a dialogue with an AI to a dialogue with a human.

[0018] Problems, configurations, effects, and the like other than those described above will become clear from the following description of the embodiments. [Brief explanation of the drawings]

[0019] [Figure 1] FIG. 1 is a diagram illustrating an example of an overview of a dialogue control system. [Figure 2] FIG. 1 illustrates an example of an embodiment of a dialogue control system. [Figure 3] FIG. 2 is a diagram illustrating an example of a hardware configuration of a dialogue control device. [Figure 4] FIG. 10 is a diagram showing an example of a dialogue switching sequence (escalation). [Figure 5] FIG. 10 illustrates an example of a system prompt. [Figure 6] FIG. 10 is a diagram showing an example of a dialogue switching sequence (operator selection). [Figure 7] FIG. 10 illustrates an example of a system prompt. DETAILED DESCRIPTION OF THE INVENTION

[0020] Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[0021] FIG. 1 is a diagram illustrating an example of an overview of a dialogue control system. In this example, the dialogue control system 1 is illustrated as being targeted at a contact center. A user (sometimes referred to as a first user) and an operator (sometimes referred to as a second user) refer to a person who requests a dialogue such as an inquiry, and a person who acts as a dialogue contact, respectively. Furthermore, the business to which the dialogue control system 1 is applied is not limited to contact centers, but may also include customer service work in general, including, for example, an in-house help desk, a corporate business contact, a hotel concierge, a museum guide, an airline cabin crew member, an avatar on a social networking service (SNS) or a video distribution system, or a character in a video game or the metaverse.

[0022] In the dialogue control system 1, a user inputs a statement and information accompanying the statement by text, voice input, image input, etc. from a user terminal 200 (sometimes referred to as a first user terminal) to a chat room (or voice call or video-based chat) provided by the dialogue control device 100 via a network 50. An interactive AI (Artificial Intelligence) of the automatic dialogue device 300 generates a response to the user terminal 200 via the network 50 and outputs it in the chat room, thereby carrying out an initial dialogue.

[0023] Here, conversational AI includes generative AI that uses a large-scale language model. The generative AI is a trained model, such as GPT-4, in which a large-scale language model using a neural network (NN) is trained by machine learning or deep learning using language data in advance. The generative AI is not limited to a neural network (NN), and other well-known methods may also be used. Furthermore, it is desirable to use a generative AI that has been tuned by few-shot learning, fine tuning, or the like.

[0024] Furthermore, the dialogue control device 100 can hand over a dialogue between a user and an interactive AI to a dialogue between the user and an operator. Specifically, the dialogue control device 100 provides a chat room in which a dialogue can take place between an operator terminal 400 (sometimes referred to as a second user terminal) used by the operator and the user terminal 200, and starts a dialogue between the user who was having a dialogue with the interactive AI and the operator, thereby escalating the dialogue from the interactive AI to the operator. Furthermore, there may be multiple operator terminals 400, and the dialogue control device 100 may select one of them when taking over the dialogue.

[0025] When escalating, the dialogue control device 100 assists the operator in understanding the history of the dialogue and starting the dialogue smoothly by presenting to the selected operator in advance handover information based on the dialogue between the user and the dialogue AI that has been processed in a predetermined manner. The handover information includes, for example, the dialogue log itself and a summary of the dialogue log. Examples of processing performed on the handover information include presentation-related processing such as sorting the dialogue log in chronological order by speaker and making statements that are closely related to the summary stand out from statements that are less relevant to the summary (for example, highlighting, graying out, or changing the font size on the dialogue log).

[0026] Furthermore, it is desirable that the user terminal 200 and the operator terminal 400 are highly portable information processing devices such as a smartphone, a PDA (Personal Digital Assistant), a PC (Personal Computer) or a tablet terminal that can communicate with the automated dialogue device 300 and the dialogue control device 100 via a network 50. The functions of the dialogue control device 100, the user terminal 200, the automated dialogue device 300 and the operator terminal 400 are realized by software that runs on an OS (Operating System).

[0027] 2 is a diagram showing an example of an embodiment of a dialogue control system. The dialogue control system 1 in this embodiment includes a dialogue control device 100, a user terminal 200 communicatively connected to the dialogue control device 100 via wired or wireless communication, an automated dialogue device 300 communicatively connected to the dialogue control device 100 via a network 50, and an operator terminal 400 communicatively connected to the dialogue control device 100 via the network 50.

[0028] The user terminal 200 is an information processing device loaned to a user by the company or organization to which the user belongs for business purposes, or is the user's own information processing device that the user carries with them. In this embodiment, the person who directly uses the user terminal 200 is a person who desires to have a conversation, such as a customer in a contact center or an employee in an in-house help desk. The operator terminal 400 is an information processing device loaned to an operator by the company or organization to which the operator belongs for business purposes, or is the operator's own information processing device that the operator carries with them. In this embodiment, the person who directly uses the operator terminal 400 is the operator.

[0029] The network 50 is, for example, the Internet or an intranet. However, the network 50 is not limited to this and may also be a WAN (Wide Area Network), a mobile phone network, or a communication network that combines these. Alternatively, the network 50 may be a VPN (Virtual Private Network) on a wireless communication network such as a mobile phone communication network.

[0030] The automated dialogue device 300 and the dialogue control device 100 are information processing devices such as so-called server devices. The automated dialogue device 300 and the dialogue control device 100 are connected to the user terminal 200 and the operator terminal 400 via a network 50.

[0031] The user terminal 200 includes at least a processing unit 220, an output unit 230, an input unit 231, and a communication unit 240. The processing unit 220 includes a dialogue processing unit 221 and realizes a so-called chat function. For example, the processing unit 220 logs into a chat room provided by the dialogue control device 100, displays a chat log, and controls browsing and input. The output unit 230 displays output information from the dialogue control device 100 or plays audio and video. The input unit 231 accepts comments input by the user in the form of text, audio, or video. The communication unit 240 communicates with other devices via wired or wireless communication. For example, the communication unit 240 communicates via various wired communication cables such as a wired LAN (Local Area Network) cable or via radio waves of a wireless LAN.

[0032] The operator terminal 400 is basically the same as the user terminal 200. The operator terminal 400 includes at least a processing unit 420, an output unit 430, an input unit 431, and a communication unit 440. The processing unit 420 includes a dialogue processing unit 421 and realizes a so-called chat function. For example, the processing unit 420 logs in to a chat room provided by the dialogue control device 100, displays a chat log (dialogue log), and controls browsing and input. The output unit 430 displays output information from the dialogue control device 100 or plays audio and video. The input unit 431 accepts comments input by the operator in the form of text, audio, or video. The communication unit 440 communicates with other devices via wired communication or wireless communication. For example, the communication unit 440 communicates via various wired communication cables such as a wired LAN cable or via radio waves of a wireless LAN.

[0033] The automated dialogue device 300 includes at least a processing unit 320, an output generation unit 330, an input unit 331, and a communication unit 340. The processing unit 320 includes an automated dialogue processing unit 321, which is a generative AI or the like that has acquired a large-scale language model, and realizes a so-called chatbot function. For example, the automated dialogue processing unit 321 logs in to a chat room provided by the dialogue control device 100, receives comments input from the user terminal 200 from the dialogue control device 100, and outputs a response to the comment to the dialogue control device 100. The automated dialogue processing unit 321 also receives and applies roles, positions, prerequisites, constraints, output conditions, instructions, etc. from the automated dialogue setting unit 123 via prompts, embedding, or an API (Application Programming Interface). The output generation unit 330 converts the output information of the automated dialogue processing unit 321 into data for delivery to the dialogue control device 100. The input unit 331 receives comments input from the dialogue control device 100. The communication unit 340 communicates with other devices via wired or wireless communication. For example, the communication unit 340 communicates via various wired communication cables such as a wired LAN cable or via radio waves of a wireless LAN.

[0034] The dialogue control device 100 includes a storage unit 110, a processing unit 120, and a communication unit 140. The storage unit 110 includes an operator list 111. At least a list including an operator as a target for escalation and contact information for dialogue (for example, chat account information, etc.) is stored.

[0035] The processing unit 120 realizes application functions. The processing unit 120 includes a dialogue control unit 121, a determination unit 122, and an automatic dialogue setting unit 123. The dialogue control unit 121 can use an automatic dialogue processing unit 321 of an automatic dialogue device 300, which is a generation AI that has acquired a large-scale language model. Specifically, the dialogue control unit 121 can use output data output by the generation AI by inputting input data to the automatic dialogue processing unit 321.

[0036] The dialogue control unit 121 provides functions for realizing dialogue (chat) between the user terminal 200 and the automatic dialogue device 300 and dialogue between the user terminal 200 and the operator terminal 400. For example, the dialogue control unit 121 sets up a chat room each time or permanently, and controls a first dialogue between the user and the automatic dialogue processing unit 321 of the automatic dialogue device, or a second dialogue between the user and the operator, to realize text chat and voice chat.

[0037] Furthermore, during the first dialogue, the dialogue control unit 121 executes dialogue handover to the second dialogue by causing the operator terminal 400 to output handover information based on the content of the first dialogue and starting the second dialogue. At this time, the dialogue control unit 121 can cause the automatic dialogue processing unit 321, which has recorded the content of the first dialogue, to generate and acquire the handover information and output it to the operator terminal 400. When outputting the handover information to the operator terminal 400, the dialogue control unit 121 can also perform a predetermined processing (processing related to the presentation described above) on the handover information and output it to the operator terminal 400.

[0038] Alternatively, the dialogue control unit 121 can record the content of the first dialogue in a recording unit (e.g., a generation AI that has acquired a large-scale language model) different from the automatic dialogue processing unit 321 as the first dialogue progresses, and can generate and acquire the handover information in the recording unit when the handover information is generated, and output it to the operator terminal 400.

[0039] Furthermore, the dialogue control unit 121 may stop the first dialogue after escalation has been performed and the second dialogue has begun. Alternatively, the dialogue control unit 121 may restrict the automatic dialogue processing unit 321 from making a comment (by issuing an instruction not to respond without permission) and allow the automatic dialogue processing unit 321 to participate in the second dialogue. In this way, the automatic dialogue processing unit 321 can incorporate the content of the second dialogue with the operator after escalation into the context.

[0040] In addition, the operator list 111 records predetermined attributes associated with each operator, and the dialogue control unit 121 causes the automatic dialogue processing unit 321 to select candidate operators who have attributes that are highly relevant to the content or summary of the first dialogue.

[0041] Furthermore, the dialogue control unit 121 can also select a successor operator from among the selected operator candidates using a model that has been trained using the performance (dialogue log) of each operator (not shown) as an explanatory variable and the user's satisfaction as a target variable. Specifically, the dialogue control unit 121 inputs the performance of each operator candidate (for example, a log or summary of a past second dialogue) into the model, accepts the operator that will achieve the highest user satisfaction as the model's output, and selects that operator.

[0042] Furthermore, the dialogue control unit 121 can also select an operator who is available (not currently in dialogue with another user) from among the selected candidate operators.

[0043] Furthermore, the dialogue control unit 121 can also display the selected operator candidates on the user terminal 200 and accept the selection of an operator from the user.

[0044] The dialogue control unit 121 can also select an operator by notifying the operator terminal 400 of the selected candidate for the operator and accepting the candidate's candidacy.

[0045] The determination unit 122 determines whether to execute handover of the dialogue from the automatic dialogue processor 321 to an operator during the first dialogue. Specifically, the determination unit 122 sets a handover occurrence condition based on the content of the first dialogue in the automatic dialogue processor 321, and determines to execute handover when the automatic dialogue processor 321 notifies the occurrence of handover. Alternatively, the determination unit 122 sets the automatic dialogue processor 321 to notify the occurrence of handover when the frequency of appearance of a predetermined keyword in the content of the first dialogue (for example, a predetermined complaint term included in a user's statement or an apology term included in a statement by the automatic dialogue processor 321) reaches a predetermined level or more. In addition, the determination unit 122 sets the automatic dialogue processor 321 to notify the occurrence of handover when the user makes a statement indicating a desire for escalation, when a situation occurs in which the automatic dialogue processor 321 is unable to provide an accurate response, when the user expresses dissatisfaction or anger with the dialogue, when a request for a decision is received from the automatic dialogue processor 321, when the volume of the user's speaking voice increases to a predetermined volume or more, when a period of silence lasts for a predetermined length or more, or the like occurs.

[0046] More specifically, the determination unit 122 determines whether any one of the following conditions or any combination thereof is met, and if so, sets the automatic dialogue processor 321 to notify that a handover has occurred. - If the first conversation contains a specific keyword - When the first user requests a takeover during the first dialogue - When it is impossible to give a response to the first user in the first dialogue (including when it is not possible to give an accurate response) - When detecting dissatisfaction with the first user's statement in the first dialogue - When the first user requests a decision in the first dialogue (a request to make a decision on behalf of the first user) When the volume of the first user's speech in the first dialogue is equal to or greater than a predetermined value - When the volume of the first user's speech during the first dialogue is below a predetermined level for a period of time that is longer than a predetermined period.

[0047] In addition, the judgment unit 122 may record the process of setting the conditions for occurrence of handover based on the first dialogue content in the automatic dialogue processing unit 321 in a recording unit for monitoring the first dialogue (a unit of an instance different from the automatic dialogue processing unit 321, for example, a generation AI with dialogue functions similar to those of the automatic dialogue processing unit 321), and have the recording unit judge the conditions for occurrence of handover based on the first dialogue content.

[0048] The automatic dialogue setting unit 123 sets dialogue parameters for the automatic dialogue processing unit 321. Specifically, the automatic dialogue setting unit 123 sets the role and position, prerequisites, constraints, output conditions, instructions, etc., of the automatic dialogue processing unit 321 in the automatic dialogue processing unit 321 by prompts, embedding, etc. The automatic dialogue setting unit 123 sets these settings for the automatic dialogue processing unit 321 via an API, etc.

[0049] The communication unit 140 communicates with other devices via wired or wireless communication. For example, the communication unit 140 communicates via various wired communication cables such as a wired LAN cable or via radio waves of a wireless LAN.

[0050] 3 is a diagram showing an example of the hardware configuration of a dialogue control device. The dialogue control device 100 has a processor 11, a memory 12, a storage 13, and a communication device 14, and each component is connected by a bus. The power source may be electricity received via a power grid, a primary battery, or a rechargeable secondary battery, and in the case of a storage battery, it can be replaceable when worn out or deteriorated.

[0051] The processor 11 is an arithmetic device such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit), and executes processing in accordance with a program recorded in the memory 12 or the storage 13. In the dialogue control device 100, processing is performed by the processor 11 operating in accordance with a program read onto the memory 12 or the storage 13. The processing unit 120, the dialogue control unit 121, the determination unit 122, and the automatic dialogue setting unit 123 each realize their respective functions by the processor 11 executing the program.

[0052] The memory 12 is a storage device such as a RAM (Random Access Memory) or a flash memory, and functions as a storage area from which programs and data are temporarily read. The storage 13 is a writable and readable storage device. The functions of the storage unit 110 are realized by the memory 12 or the storage 13. Note that the functions of the storage unit 110 may also be realized by a storage device connected via the communication device 14.

[0053] The communication device 14 is an interface for connecting the dialogue control device 100 to an external device for communication. For example, the communication device 14 performs wireless communication using an antenna that can use a predetermined radio wave (e.g., 5 GHz band, 2.4 GHz band, etc.) to establish a connection with the dialogue control device 100 according to the Wi-Fi standard.

[0054] The processing of each component of the dialogue control device 100 may be executed by one piece of hardware or by multiple pieces of hardware. Furthermore, the processing of each component of the dialogue control device 100 may be realized by one program or by multiple programs.

[0055] The user terminal 200, the automated dialogue device 300, and the operator terminal 400 have roughly the same hardware configuration as the dialogue control device 100. However, the user terminal 200 and the operator terminal 400 often include input / output devices such as a touch panel display and a microphone, similar to a typical smartphone.

[0056] Next, the processing flow of the dialogue control system 1 will be described. FIG. 4 is a diagram showing an example of a dialogue switching sequence (escalation) of the dialogue control system 1. The processing of this sequence diagram covers a sequence in which a chat room is opened in response to a request from the user terminal 200, a dialogue is conducted between the automated dialogue device 300 and the user in the chat, and the dialogue is escalated to a dialogue with an operator. Also, depending on the specifications of the chat room, some are opened at any time, while others are opened once and maintained semi-permanently. However, this sequence includes a situation in which a chat room is opened for the first time. Therefore, from the second time onwards, the chat room may already be opened, or may be newly opened.

[0057] First, the user terminal 200 issues a dialogue start instruction to the dialogue control device 100 (step S01). Specifically, the user terminal 200 transmits an instruction to open a chat room.

[0058] Then, the dialogue control device 100 transmits the automatic dialogue setting to the automatic dialogue device 300 (step S02). Specifically, the automatic dialogue setting unit 123 passes the parameters and prompts used in the automatic dialogue to the automatic dialogue processing unit 321 via the API.

[0059] Then, while participating in the chat, the user makes a statement as an input to the dialogue in the form of text, voice, or video via the dialogue processing unit 221 of the user terminal 200 (step S03). The dialogue processing unit 221 transmits the inputted statement to the dialogue control device 100 as text information, voice, or video on the chat.

[0060] Then, the dialogue control unit 121 accepts the utterances transmitted as text information, audio, or video on the chat, and transfers the dialogue input to the automatic dialogue device 300 (step S04). Specifically, the dialogue control unit 121 accepts the utterances transmitted as text information, audio, or video on the chat, and transfers the dialogue input to the automatic dialogue processing unit 321 via an API or as text data on a prompt.

[0061] Then, the automatic dialogue processing unit 321 of the automatic dialogue device 300 outputs a utterance as an output of the automatic dialogue in the form of text, voice, or video (step S05). Specifically, the automatic dialogue processing unit 321 transmits the automatic dialogue output to the dialogue control device 100 as text information, voice, or video on a chat.

[0062] Then, the dialogue control unit 121 accepts automatic dialogue output, which is a statement sent as text information, audio, or video on the chat, and transfers it as dialogue output to the user terminal 200 (step S06). Specifically, the dialogue control unit 121 accepts a statement sent as text information, audio, or video on the chat, and transfers it as dialogue output to the user terminal 200 via an API or as text data on a prompt.

[0063] In this way, the processes of steps S03 to S06 constitute a series of dialogues, and a first dialogue is formed by carrying out a plurality of dialogues in succession.

[0064] Then, for example, when at least one of the conditions for handover is satisfied and a handover occurrence notification is received from the automatic dialogue processing unit 321 (step S07), the judgment unit 122 determines that the handover will be performed and instructs the automatic dialogue processing unit 321 (or a recording unit for monitoring the first dialogue) to perform the handover (step S08).

[0065] When the automatic dialogue processing unit 321 (or the recording unit for monitoring the first dialogue) is instructed to take over, it generates handover information and transmits it to the dialogue control device 100 (step S09). Specifically, the automatic dialogue processing unit 321 (or the recording unit for monitoring the first dialogue) generates the dialogue log of the first dialogue itself, a summary of the dialogue log, etc.

[0066] The dialogue control unit 121 performs predetermined processing on the transmitted handover information and transmits it to the operator terminal 400 of one of the operators included in the operator list 111 for display (step S10). Examples of processing performed on the handover information include processing related to presentation, such as sorting the dialogue logs in chronological order by speaker, and processing to make statements that are more relevant to the summary stand out from statements that are less relevant to the summary (for example, highlighting, graying out, or changing the font size on the dialogue log).

[0067] The dialogue processing unit 421 of the operator terminal 400 participates in a chat with the user, and the operator makes a statement as an input for the dialogue using text, voice, or video (step S11). The dialogue processing unit 421 transmits the inputted statement to the dialogue control device 100 as text information, voice, or video on the chat.

[0068] Then, the dialogue control unit 121 accepts the utterance transmitted as text information, audio, or video on the chat, and transfers the utterance as dialogue output to the user terminal 200 (step S12). Specifically, the dialogue control unit 121 accepts the utterance transmitted as text information, audio, or video on the chat, and transfers the dialogue output to the dialogue processing unit 221 via the API or as text information, audio, or video on the chat.

[0069] Then, the user makes a statement as an input for the dialogue using text, voice, or video via the dialogue processing unit 221 of the user terminal 200 (step S13). The dialogue processing unit 221 transmits the input statement to the dialogue control device 100 as text information, voice, or video on the chat.

[0070] Then, the dialogue control unit 121 accepts the utterance transmitted as text information, audio, or video on the chat, and transfers the utterance as dialogue output to the operator terminal 400 (step S14). Specifically, the dialogue control unit 121 accepts the utterance transmitted as text information, audio, or video on the chat, and transfers the utterance to the automatic dialogue processing unit 321 via the API. Alternatively, the dialogue control unit 121 transfers the utterance to the automatic dialogue processing unit 321 as text data on a prompt.

[0071] In this way, the processes of steps S11 to S14 constitute a series of dialogues, and a second dialogue is formed by carrying out a plurality of dialogues in succession.

[0072] The above is an example of a dialogue switching sequence (escalation). According to this example of a dialogue switching sequence (escalation), a seamless escalation can be achieved by handing over a dialogue with an AI to a dialogue with a human.

[0073] 5 is a diagram showing an example of a system prompt. System prompt 800 is an example of a summary of a prompt given to the generation AI of the automatic dialogue processor 321 to cause it to execute handover to an operator. System prompt 800 includes instructions such as the role and instructions to create handover information from the dialogue log (a summary of up to 200 characters and highlights on the dialogue log).

[0074] 6 is a diagram showing an example of a dialogue switching sequence (operator selection) of the dialogue control system 1. The dialogue switching sequence (operator selection) is basically the same processing as the dialogue switching sequence (escalation), but it embodies a process in which the automatic dialogue processing unit 321 of the automatic dialogue device 300 or the recording unit selects a candidate when the dialogue control unit 121 selects an operator to be the target of escalation execution.

[0075] In step S09, the automatic dialogue processing unit 321 (or a recording unit for monitoring the first dialogue) generates handover information and transmits it to the dialogue control device 100. The dialogue control unit 121 of the dialogue control device 100 causes the automatic dialogue processing unit 321 (or a recording unit for monitoring the first dialogue) to select one or more second user candidates to engage in the second dialogue from a list (operator list 111) containing multiple second users, according to the content of the first dialogue included in the handover information (step S60). Here, the dialogue control unit 121 may cause the automatic dialogue processing unit 321 to select operator candidates having attributes that are highly related to the content or summary of the first dialogue.

[0076] The automatic dialogue processing unit 321 (or a recording unit for monitoring the first dialogue) selects one or more second user candidates (operator candidates) who will engage in the second dialogue from a list (operator list 111) containing multiple second users, according to the content of the first dialogue included in the handover information, and transmits the selected candidates to the dialogue control device 100 (step S61). Here, the automatic dialogue processing unit 321 (or a recording unit for monitoring the first dialogue) may select operator candidates who have attributes that are highly relevant to the content or summary of the first dialogue.

[0077] Then, the dialogue control unit 121 selects one operator from the operator candidates (step S62). Here, the dialogue control unit 121 can select a successor operator from the selected operator candidates using a model that has been trained using the performance (dialogue log) of each operator (not shown) as an explanatory variable and the user's satisfaction as a target variable. The dialogue control unit 121 can also select an operator who is available (not currently in dialogue with another user) from the selected operator candidates. The dialogue control unit 121 can also display the selected operator candidates on the user terminal 200 and accept the user's selection of an operator. The dialogue control unit 121 can also select an operator by notifying the operator terminal 400 of the selected operator candidate and accepting the candidacy.

[0078] Then, the dialogue control unit 121 processes the handover information received in step S09 (step S40). Examples of processing the handover information include processing related to presentation, such as sorting the dialogue logs in chronological order by speaker, and processing to make statements that are more relevant to the summary stand out from statements that are less relevant to the summary (for example, highlighting, graying out, or changing the font size on the dialogue log).

[0079] In this way, when the dialogue control unit 121 selects an operator to be the target of escalation, the dialogue control unit 121 causes the automatic dialogue processing unit 321 of the automatic dialogue device 300 or the recording unit to select candidates and starts a second dialogue with one operator selected from the candidates.

[0080] The above is an example of the dialogue switching sequence (operator selection). According to the dialogue switching sequence (operator selection), an operator can be automatically selected at the time of escalation, making it easier to escalate from a chatbot or the like.

[0081] FIG. 7 is a diagram showing an example of a system prompt. System prompt 700 is an example of a summary of a prompt given to the generation AI of the automatic dialogue processor 321 to select a candidate operator to take over. System prompt 700 includes an instruction to read the operator list 111 and select a candidate operator, including the role and conditions for selecting an operator. The instruction to select a candidate operator is given by passing the operator list 111 to the generation AI of the automatic dialogue processor 321 to make a selection. However, the example is not limited to this, and it is also possible to limit the instruction to notifying the characteristics of the operator to take over.

[0082] Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described exemplary embodiments and includes various modifications. For example, in the above-described embodiment, the automatic dialogue processor 321 and the operator are recognizable as different entities by the user, but this is not limiting. For example, even if the entity generating the utterances is switched due to escalation, the automatic dialogue processor 321 and the operator may use the same avatar (or account). In this way, the entity generating the utterances can be switched without the user being aware of the switch in the entity.

[0083] Alternatively, in the above embodiment, escalation is performed between the automatic dialogue processor 321 and the operator, but this is not limited to this. For example, multiple automatic dialogue processors 321 using different information for learning or different algorithms may be provided, and escalation may be performed between the automatic dialogue processors 321 depending on the type of dialogue content (such as a standard explanation of a service, a more complex introduction of a necessary service based on the user's concerns, or the difficulty of acquiring specialized knowledge required for the dialogue). In this way, efficient automatic dialogue is possible.

[0084] The above-described exemplary embodiments have been described in detail to facilitate understanding of the present invention, and the present invention is not limited to those including all of the components described herein. Furthermore, it is possible to replace part of the configuration of one embodiment with the configuration of another embodiment. It is also possible to add the configuration of another embodiment to the configuration of one embodiment. It is also possible to add, delete, or replace part of the configuration of each embodiment with other configurations. Furthermore, some or all of the above-described configurations, functions, processing units, processing means, etc. may be realized in hardware, for example, by designing them as integrated circuits. Furthermore, the control lines and information lines in the figures are those considered necessary for explanation, and are not necessarily all shown. It is also possible to consider that almost all of the components are interconnected. [Explanation of symbols]

[0085] 1 dialogue control system, 50 network, 100 dialogue control device, 110 storage unit, 111 operator list, 120 processing unit, 121 dialogue control unit, 122 determination unit, 123 automatic dialogue setting unit, 140 communication unit, 200 user terminal, 220 processing unit, 221 dialogue processing unit, 230 Output unit, 231···input unit, 240···communication unit, 300···automated dialogue device, 320···processing unit, 321···automatic dialogue processing unit, 330···output generation unit, 331···input unit, 340···communication unit, 400···operator terminal, 420···processing unit, 421···dialogue processing unit, 430···output unit, 431···input unit, 440···communication unit.

Claims

1. A dialogue control device connectable to each of a first user terminal, a plurality of second user terminals, and a dialogue AI using a large-scale language model, A dialogue control unit that controls a first dialogue between a first user using the first user terminal and the dialogue AI, and a second dialogue between the first user and a second user using any of the second user terminals; a determination unit that determines whether or not to execute handover to any of the second dialogues during the first dialogue, When the determination unit determines to execute handover during the first dialogue, the dialogue control unit causes the dialogue AI that recorded the content of the first dialogue to select one or more candidates for the second user to engage in the second dialogue from a list including a plurality of the second users according to the content of the first dialogue, and starts the second dialogue with one of the second users selected from the candidates. A dialogue control device characterized by:

2. 2. A dialogue control device according to claim 1, a predetermined attribute is associated with each of the second users in the list; The dialogue control unit causes the dialogue AI to select candidates for the second user who have the attributes highly related to the summary of the first dialogue. A dialogue control device characterized by:

3. 2. A dialogue control device according to claim 1, the dialogue control unit selects the second user from the selected candidates for the second user using a model trained using the performance of the second user as an explanatory variable and the satisfaction level of the first user as an objective variable. A dialogue control device characterized by:

4. 2. A dialogue control device according to claim 1, the interaction control unit selects the second user who is available to respond from among the selected second user candidates; A dialogue control device characterized by:

5. 2. A dialogue control device according to claim 1, the interaction control unit displays the selected candidates for the second user on the first user terminal and accepts a selection of the second user from the first user; A dialogue control device characterized by:

6. 2. A dialogue control device according to claim 1, the dialogue control unit notifies the second user terminal of the selected candidate for the second user and selects the second user by accepting the candidacy. A dialogue control device characterized by:

7. A dialogue control system including a dialogue control device connectable to each of a first user terminal, a plurality of second user terminals, and a dialogue AI using a large-scale language model, A dialogue control step of controlling a first dialogue between a first user using the first user terminal and the dialogue AI, and a second dialogue between the first user and a second user using any of the second user terminals; a determination step of determining whether or not to execute a handover to any of the second dialogues during the first dialogue; In the dialogue control step, when it is determined in the determination step that handover is to be performed during the first dialogue, the dialogue AI that recorded the content of the first dialogue selects one or more candidates for the second user to engage in the second dialogue from a list including a plurality of the second users according to the content of the first dialogue, and starts the second dialogue with one of the second users selected from the candidates. A dialogue control system characterized by:

8. A dialogue control method using a dialogue control device connectable to each of a first user terminal, a plurality of second user terminals, and a dialogue AI using a large-scale language model, the method comprising: The dialogue control device includes a processing unit, The processing unit A dialogue control step of controlling a first dialogue between a first user using the first user terminal and the dialogue AI, and a second dialogue between the first user and a second user using any of the second user terminals; a determination step of determining whether or not to execute a handover to any of the second dialogues during the first dialogue; In the dialogue control step, when it is determined in the determination step that handover is to be performed during the first dialogue, the dialogue AI that recorded the content of the first dialogue selects one or more candidates for the second user to engage in the second dialogue from a list including a plurality of the second users according to the content of the first dialogue, and starts the second dialogue with one of the second users selected from the candidates. A dialogue control method comprising:

9. A dialogue control program that causes an information processing device to control a dialogue, The information processing device includes a processing unit, The processing unit A dialogue control step of controlling a first dialogue between a first user using a first user terminal and a dialogue AI using a large-scale language model, and a second dialogue between the first user and a second user using one of a plurality of second user terminals; a determination step of determining whether or not to execute handover to any of the second dialogues during the first dialogue; In the dialogue control step, when it is determined in the determination step that handover is to be performed during the first dialogue, the dialogue AI that recorded the content of the first dialogue selects one or more candidates for the second user to engage in the second dialogue from a list including a plurality of the second users according to the content of the first dialogue, and starts the second dialogue with one of the second users selected from the candidates. A dialogue control program characterized by:

Images