Conversation providing device

Abstract

Provided is a conversation providing device capable of improving motivation for conversation with a virtual agent using artificial intelligence.　The conversation providing device comprises a conversation unit (32) that instructs a task processing device (60), which processes tasks using artificial intelligence in response to a task processing instruction, to generate conversation content of a virtual agent with respect to a user, acquires the conversation content generated by the task processing device (60), and outputs the conversation content to a user terminal device. The conversation unit (32) instructs the task processing device (60) to generate conversation content corresponding to a topic set from among topic candidates in accordance with weights that are assigned to the topic candidates on the basis of a weighting viewpoint.

Description

Conversation Providing Device Cross-Reference to Related Applications

[0001] This application claims the benefit of Japanese Patent Application No. 2025-066512 filed in Japan on April 14, 2025, the content of which is incorporated herein by reference in its entirety.

[0002] The present invention relates to a conversation providing device that provides a user with a conversation with a virtual agent using artificial intelligence.

[0003] Conversations with virtual agents using artificial intelligence are being provided to users. In generating such conversations of a virtual agent, a predetermined mode is selected from a plurality of modes, and topics are specified according to the selected mode (for example, refer to Japanese Patent No. 7530688).

[0004] When continuing a conversation with a virtual agent using artificial intelligence, similar conversations will be repeated from the virtual agent, which reduces the motivation for the conversation with the virtual agent.

[0005] An object of the present invention is to provide a conversation providing device capable of improving the motivation for conversation with a virtual agent using artificial intelligence.

[0006] One aspect of the present invention is a conversation providing device comprising a conversation unit that instructs a task processing device, which processes tasks using artificial intelligence in response to a task processing instruction, to generate conversation content of a virtual agent for a user, acquires the conversation content generated by the task processing device, and outputs the acquired conversation content to a user terminal device, wherein the conversation unit instructs the task processing device to generate conversation content corresponding to a topic set from among each of the topic candidates according to a weight assigned to each of the topic candidates based on a weighting perspective.

[0007] The present invention makes it possible to improve the motivation for conversation with a virtual agent using artificial intelligence.

[0008] A block diagram showing the hardware configuration of a conversation provision system according to one embodiment of the present invention. A block diagram showing the functional configuration of a conversation provision system according to one embodiment of the present invention. A first sequence diagram showing a conversation provision method according to one embodiment of the present invention. A second sequence diagram showing a conversation provision method according to one embodiment of the present invention. A third sequence diagram showing a conversation provision method according to one embodiment of the present invention. A fourth sequence diagram showing a conversation provision method according to one embodiment of the present invention. A fifth sequence diagram showing a conversation provision method according to one embodiment of the present invention. A table diagram illustrating newly acquired information according to one embodiment of the present invention. A table diagram illustrating how to set the type of prompt according to one embodiment of the present invention. A table diagram illustrating the topic and weighting perspective of an information provision prompt according to one embodiment of the present invention. A table diagram illustrating the topic and weighting perspective of an information collection prompt according to one embodiment of the present invention.

[0009] An embodiment of the present invention will be described with reference to Figures 1 to 7. Hereinafter, a virtual agent that converses with a user using artificial intelligence will be referred to as an AI agent. Such AI agents may provide information to the user according to a certain purpose or assist the user in performing various procedures. However, an AI agent that primarily converses with the user is expected not only to provide information and perform various procedures according to a certain purpose, but also to build a close relationship with the user, gain the user's trust, and become a supportive presence for the user.

[0010] In this embodiment, the conversation content of the AI ​​agent to the user is determined by setting a predetermined conversation direction from among multiple conversation directions based on predefined conditions, then assigning weights to each candidate topic based on a weighting perspective corresponding to the conversation direction, and generating conversation content corresponding to the set topic from among the candidate topics according to the weights assigned to each candidate topic. This prevents the AI ​​agent from repeating similar conversations even as the conversation continues, improves the motivation to converse with the AI ​​agent, and makes it possible to build a close relationship between the user and the AI ​​agent.

[0011] As shown in Figures 1 and 2, the conversation provision system of this embodiment includes a conversation provision server 10, a generation server 60, and a user terminal 70.

[0012] Referring to Figure 1, the hardware configuration of the conversation provision system in this embodiment will be described. Since the hardware configurations of the conversation provision server 10 and the generation server 60 are the same, they will be described together as servers below.

[0013] As shown in Figure 1, servers 10 and 60 are physically configured as computers including a processor 211, memory 212, storage 213, communication device 214, input device 215, output device 216, and buses connecting them. Each of these devices operates on power supplied from a battery (not shown). In the following description, the term "device" can be read as a circuit, device, unit, etc. The hardware configuration of servers 10 and 60 may include one or more of the devices shown in the figure, or it may be configured without some of the devices. Alternatively, multiple devices with different enclosures may be connected via communication to constitute servers 10 and 60.

[0014] Each function in servers 10 and 60 is realized by loading predetermined software (programs) onto hardware such as the processor 211 and memory 212, which causes the processor 211 to perform calculations, control communication by the communication device 214, and control at least one of the reading and writing of data in the memory 212 and storage 213.

[0015] The processor 211 controls the entire computer, for example, by running the operating system. The processor 211 may be composed of a central processing unit (CPU) that includes interfaces with peripheral devices, control devices, arithmetic units, registers, etc. Also, for example, a baseband signal processing unit, a call processing unit, etc., may be implemented by the processor 211.

[0016] The processor 211 reads programs (program code), software modules, data, etc., from at least one of the storage 213 and the communication device 214 into the memory 212 and executes various processes accordingly. The program used is one that causes the computer to execute at least a part of the operations described later. Functional blocks of servers 10 and 60 are stored in the memory 212 and may be implemented by control programs that run on the processor 211. Various processes may be executed by one processor 211, or they may be executed simultaneously or sequentially by two or more processors 211. The processor 211 may be implemented by one or more chips. The program may also be transmitted to the servers 10 and 60 via a telecommunications line.

[0017] The memory 212 is a computer-readable recording medium and may consist of at least one of the following: ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), RAM (Random Access Memory), etc. The memory 212 may also be referred to as a register, cache, main memory, etc. The memory 212 can store executable programs (program code), software modules, etc., for carrying out the method according to this embodiment.

[0018] The storage 213 is a computer-readable recording medium and may consist of at least one of the following: an optical disc such as a CD-ROM (Compact Disc ROM), a hard disk drive, a flexible disk, a magneto-optical disk (e.g., a compact disc, a digital multipurpose disc, a Blu-ray® disc), a smart card, flash memory (e.g., a card, a stick, a key drive), a floppy® disk, a magnetic strip, etc. The storage 213 may also be referred to as an auxiliary storage device.

[0019] The communication device 214 is hardware (transceiver / receiver device) for communicating between computers via at least one of a wired network and a wireless network, and is also referred to as a network device, network controller, network card, communication module, etc.

[0020] Each device, such as the processor 211 and memory 212, is connected by a bus for communicating information. The bus may be configured using a single bus, or different buses may be used for each device.

[0021] Servers 10 and 60 may be configured to include hardware such as microprocessors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable logic devices (PLDs), and field-programmable gate arrays (FPGAs), and some or all of each functional block may be realized by such hardware. For example, processor 211 may be implemented using at least one of these hardware components.

[0022] The user terminal 70 is a computer such as a smartphone, mobile phone, tablet, or wearable device. Physically, the user terminal 70 is configured as a computer device including a processor 261, memory 262, storage 263, communication device 264, input device 265, output device 266, and a bus connecting these. The processor 261, memory 262, and storage 263 are hardware similar to the processor 211, memory 212, and storage 213 of servers 10 and 60.

[0023] The communication device 264 may include, for example, a high-frequency switch, duplexer, filter, frequency synthesizer, etc., to implement at least one of frequency division duplex (FDD) and time division duplex (TDD). For example, the transmitting and receiving antennas, amplifier section, transmitting and receiving section, transmission path interface, etc., may be implemented by the communication device 264. The transmitting and receiving section may be physically or logically separated into a transmitting section and a receiving section.

[0024] The input device 265 is an input device that accepts input from an external source (e.g., a key, microphone, switch, button, sensor, reader, scanner, etc.). The output device 266 is an output device that outputs to an external source (e.g., a display, speaker, LED lamp, etc.). The input device 265 and the output device 266 may be configured as an integrated unit (e.g., a touchscreen).

[0025] Referring to Figures 2 and 3, the functional configuration and conversation provision method of this embodiment will be described. Hereinafter, each functional configuration of the conversation provision system will be described in accordance with each step of the conversation provision method. The conversation provision program of this embodiment works in cooperation with the hardware configuration of the conversation provision system of this embodiment described above to realize each functional configuration of the conversation provision system of this embodiment and to execute each step of the conversation provision method.

[0026] Furthermore, in one embodiment described below, the user is provided with a conversation with the AI ​​agent through voice input and voice output, but the user may also be provided with a conversation with the AI ​​agent through text input and text output on the screen.

[0027] As shown in Figure 2, in the conversation provision system, the conversation provision server 10 stores the following information in the user database 12 (hereinafter also referred to as "user DB 12"): user identification information that identifies the AI ​​agent that engages in conversation with the user, conversation history information that shows the history of conversation content between the user and the AI ​​agent, conversation status information that shows the most recent conversation status between the user and the AI ​​agent, familiarity information that shows the familiarity level of the user with the AI ​​agent, positivity information that shows the degree of positivity of the conversation content of the most recent conversation between the user and the AI ​​agent, personal information that shows the user's own personal information, location history information that shows the history of the user's current location, and schedule history information that shows the history of the user's schedule. In this embodiment, the conversation status information includes information provision status, which is a conversation status in which it is appropriate for the AI ​​agent to provide topics to the user, and information collection status, which is a conversation status in which it is appropriate for the AI ​​agent to ask questions to the user. Personal information includes numerous items, such as the user's address and name (so-called personal information), as well as occupational information indicating the user's profession, preference information indicating the user's tastes, and health information indicating the user's health status.

[0028] Furthermore, the agent database 14 (hereinafter also referred to as "Agent DB 14") stores agent feature information for numerous AI agents, associated with each agent's identification information, indicating the characteristics of the AI ​​agent. The agent feature information includes character feature information indicating the characteristics of the AI ​​agent's character, and voice information for generating the AI ​​agent's voice.

[0029] As shown in Figures 2 and 3A, the conversation provision system and conversation provision method allow the user to select an AI agent to engage in conversation with, as described below.

[0030] Conversation Application Startup Step S12: The user terminal 70 has a conversation application installed for having a conversation with the AI ​​agent. In the user terminal 70, the application startup unit 72 starts the conversation application in response to the user's operation to start the conversation application.

[0031] Agent Selection Step S14: In the user terminal 70, the agent selection unit 74 presents the user with a number of AI agents prepared in the conversation application and allows the user to select an AI agent to converse with. The user can select an AI agent through on-screen selection, voice input, etc. The number of AI agents prepared in the conversation application corresponds to existing characters, original characters of the conversation application, etc., and the user can select their favorite character to be the AI ​​agent that converses with them.

[0032] Agent designation request step S16: On the user terminal 70, the agent selection unit 74 sends an agent designation request to the conversation provision server 10, along with user identification information to identify the user and agent identification information to identify the AI ​​agent, to designate the AI ​​agent selected by the user.

[0033] Agent designation step S18: In the conversation provision server 10, the agent designation unit 16 stores the agent identification information in the user DB 12 in association with the user identification information, in response to the agent designation request received from the user terminal 70 along with the user identification information and agent identification information.

[0034] Agent image display step S22: On the user terminal 70, the image generation unit 76 generates an agent image representing the AI ​​agent, and the image display unit 78 displays the agent image generated by the image generation unit 76 to the user.

[0035] As shown in Figures 2 and 3B, the conversation provision system and conversation provision method set or update conversation status information, intimacy information, positivity information, and personal information in the user DB 12, as described below.

[0036] In the conversation analysis instruction step S26, the conversation analysis instruction unit 18 in the conversation provision server 10 extracts conversation history information from the user DB 12, which shows the history of conversation content between the user and the AI ​​agent, associated with the user identification information. Based on the extracted conversation history information, the conversation analysis instruction unit 18 generates a conversation situation extraction prompt, a sensitivity level information extraction prompt, a positivity level information extraction prompt, and a personal information extraction prompt, and sends them to the generation server 60 along with the user identification information. The conversation situation extraction prompt, sensitivity level information extraction prompt, positivity level information extraction prompt, and personal information extraction prompt respectively instruct the generation AI for conversation analysis to extract conversation situation information showing the most recent conversation situation between the user and the AI ​​agent, sensitivity level information showing the sensitivity level of the conversation content between the user and the AI ​​agent, positivity level information showing the positivity level of the user's most recent conversation content with the AI ​​agent, and personal information showing the user's personal information, from the conversation history information. In this embodiment, the conversation situation information extracted is either an information provision situation, where it is appropriate for the AI ​​agent to provide topics to the user, or an information gathering situation, where it is appropriate for the AI ​​agent to ask questions to the user. Personal information includes a large number of items, such as so-called personal information such as the user's address and name, as well as occupational information indicating the user's occupation, preference information indicating the user's preferences, and health status information indicating the user's health status. The conversation analysis instruction unit 18 also outputs the extracted conversation history information along with user identification information to the intimacy setting unit 24.

[0037] In the conversation analysis step S28, the generation server 60 has a conversation analysis unit 62 equipped with a generation AI for conversation analysis that utilizes LLM (Large Language Models). The conversation analysis unit 62 inputs the conversation situation extraction prompt, sensitivity level information extraction prompt, positivity level information extraction prompt, and personal information extraction prompt, which it has received from the conversation provision server 10 along with user identification information, into the generation AI for conversation analysis. The conversation analysis unit 62 receives the output of conversation situation information, sensitivity level information, positivity level information, and personal information from the generation AI for conversation analysis, and transmits the conversation situation information, sensitivity level information, positivity level information, and personal information to the conversation provision server 10 along with user identification information.

[0038] In the conversation status setting step S32, the conversation status setting unit 22 of the conversation provision server 10 sets the conversation status information based on the conversation status information received from the generation server 60 along with the user identification information, and stores the conversation status information in the user DB 12 in association with the user identification information.

[0039] In the intimacy setting step S34, the conversation provision server 10 sets intimacy information indicating the intimacy level of the user identified by the user identification information towards the AI ​​agent, based on the amount of conversation indicated by the conversation history information input from the conversation analysis instruction unit 18 along with the user identification information, and the sensitivity level of the conversation indicated by the sensitivity level information received from the generation server 60 along with the user identification information, and stores the intimacy information in the user DB 12 in association with the user identification information.

[0040] Regarding the setting of intimacy, instead of analyzing conversation history using a generative AI for conversation analysis, it may be set based on the amount of conversation between the user and the AI ​​agent, or it may be set based on the number of times, frequency, and total time the AI ​​agent has been active, or these may be combined with the analysis of conversation history using a generative AI for conversation analysis.

[0041] Positivity degree setting step S36 In the conversation providing server 10, the positivity degree setting unit 26 sets the positivity degree information based on the positivity degree information received together with the user identification information from the generation server 60, and stores the positivity degree information in the user DB 12 in association with the user identification information.

[0042] Personal information setting step S38 In the conversation providing server 10, the personal information setting unit 28 sets the personal information based on the personal information received together with the user identification information from the generation server 60, and stores the personal information in the user DB 12 in association with the user identification information.

[0043] For the setting of personal information, instead of analyzing the conversation history by generative AI for conversation analysis, the setting may be performed by the user's own input, or by setting preference information indicating the user's preference based on the user's purchase history on the user terminal 70, content viewing or browsing history, search and browsing history on websites, etc., or may be set by combining these with the analysis of the conversation history by generative AI for conversation analysis.

[0044] Update step S42 When the conversation history information is updated in the user DB 12, the conversation status information, familiarity information, positivity degree information, and personal information in the user DB 12 are updated in the same manner as the above steps.

[0045] As shown in Figure 2, Figure 3C and Figure 3D, in the conversation providing system and the conversation providing method, a conversation is conducted between a user and an AI agent as described below.

[0046] Related information collection step S46 In the user terminal 70, the related information collection unit 82 collects related information for generating a conversation of the AI agent. In the present embodiment, the related information includes new arrival notification related information related to new arrival notifications from other applications of the user terminal 70, current position information indicating the current position of the user collected from a position information application, and schedule information indicating the user's schedule collected from a schedule management application.

[0047] Utterance generation request step S48 At the user terminal 70, the conversation execution unit 84 transmits an utterance generation request for requesting generation of utterance content of the AI agent's utterance to the user, together with user identification information and related information, to the conversation providing server 10.

[0048] Utterance generation instruction step S52 In the conversation providing server 10, the conversation generation instruction unit 32 serving as a conversation unit extracts agent identification information and conversation history information associated with the user identification information from the user DB 12 in response to the utterance generation request received together with the user identification information and the related information from the user terminal 70, and extracts character feature information of the agent feature information associated with the agent identification information from the agent DB 14. When the conversation history information is large in volume, only a fixed amount of the most recent conversation history information is extracted. The character feature information indicates the character features of the AI agent.

[0049] Subsequently, the conversation generation instruction unit 32 generates an utterance generation prompt based on the conversation history information and the character feature information, and transmits the utterance generation prompt together with the user identification information to the generation server 60. The utterance generation prompt is for instructing generative AI for conversation generation to generate utterance text data indicating the utterance content of the AI agent's utterance to the user, and gives an instruction such that the utterance content of the AI agent is natural in view of the conversation history and matches the character features of the AI agent. The method for generating the utterance generation prompt will be described in detail later.

[0050] Utterance generation step S54 In the generation server 60, the conversation generation unit 64 is provided with generative AI for conversation generation that uses an LLM. The conversation generation unit 64 inputs the utterance generation prompt received together with the user identification information from the conversation providing server 10 to the generative AI for conversation generation, receives the output of utterance text data indicating the utterance content of the AI agent's utterance from the generative AI for conversation generation, and transmits the utterance text data together with the user identification information to the conversation providing server 10.

[0051] In the conversation history update step S56, the conversation generation instruction unit 32 in the conversation provision server 10 updates the conversation history information by adding the utterance text data received from the generation server 60 along with the user identification information to the conversation history information associated with the user identification information in the user DB 12. The conversation generation instruction unit 32 also outputs the utterance text data along with the user identification information to the speech generation unit 46.

[0052] Speech voice generation step S58: In the conversation provision server 10, when the speech generation unit 46 receives speech text data along with user identification information from the conversation generation instruction unit 32, it extracts agent identification information associated with the user identification information from the user DB 12, extracts voice information of agent feature information associated with the agent identification information from the agent DB 14, generates speech voice data by converting the speech text data into speech using the voice of the AI ​​agent, and transmits it to the user terminal 70 identified by the user identification information.

[0053] Speech Voice Output Step S62: In the user terminal 70, the conversation execution unit 84 acquires the speech voice data received from the conversation provision server 10 and outputs it to the voice output unit 86. Based on the speech voice data input from the conversation execution unit 84, the voice output unit 86 outputs the AI ​​agent's speech to the user as voice.

[0054] In the user voice acquisition step S64, the voice acquisition unit 88 in the user terminal 70 acquires the voice of the conversation between the user and the AI ​​agent and obtains user voice information indicating the user's voice.

[0055] In the response generation request step S66, the conversation execution unit 84 of the user terminal 70 sends a response generation request to the conversation provision server 10, requesting the generation of a response content for the AI ​​agent's response to the user, along with the user voice information and user identification information acquired by the voice acquisition unit 88.

[0056] In the user voice analysis step S68, the voice analysis unit 48 of the conversation provision server 10 analyzes the user voice information received from the user terminal 70 along with the user identification information, generates user voice text data by converting the user voice into text, and outputs it to the conversation generation instruction unit 32 along with the user identification information.

[0057] In the conversation history update step S72, the conversation generation instruction unit 32, when it receives user voice text data along with user identification information from the voice analysis unit 48, updates the conversation history information in the user DB 12 by adding the user voice text data to the conversation history information associated with the user identification information.

[0058] In response generation instruction step S74, the conversation generation instruction unit 32 in the conversation provision server 10 extracts agent identification information and conversation history information associated with user identification information from the user DB 12, and extracts character feature information of agent feature information associated with agent identification information from the agent DB 14.

[0059] Next, the conversation generation instruction unit 32 generates a response generation prompt based on the user voice text data, conversation history information, and character feature information, and sends it to the generation server 60 along with the user identification information. The response generation prompt instructs the conversation generation AI to generate response text data that shows the content of the AI ​​agent's response to the most recent conversation from the user indicated by the user voice text data. The instruction is to ensure that the AI ​​agent's response is natural in light of the conversation history and matches the character features of the AI ​​agent.

[0060] The subsequent steps are the same as those described above for generating the content of the AI ​​agent's utterances to the user.

[0061] As shown in Figures 2 and 3E, the conversation provision system and conversation provision method generate speech generation prompts as described below.

[0062] As described above, in the conversation provision server 10, the conversation generation instruction unit 32 generates a speech generation prompt that instructs the conversation generation AI to generate speech text data indicating the content of the AI ​​agent's utterance to the user. The method for generating the speech generation prompt is as follows.

[0063] In the new information acquisition step S78, the conversation generation instruction unit 32 acquires new information, which is new information for the user.

[0064] In this embodiment, for example, as shown in Figure 4, new information is acquired as follows: (1) New notification related information If the related information received from the user terminal 70 includes new notification related information, the new notification related information is acquired as new information.

[0065] (2) Extract user preference information associated with user identification information from the news-related information user DB12, and search for news-related information such as news articles and releases that match the user's preferences indicated by the user preference information using internet search, etc. If new news-related information is found, acquire it as new information.

[0066] (3) Location-related information: Location history information associated with user identification information is extracted from the user DB 12. If the user's current location, as indicated by the current location information received from the user terminal 70, is not included in the user's current location history indicated by the location history information, location-related information regarding the user's current location and its surroundings is searched for by internet search, etc. If new location-related information is extracted, it is acquired as new information.

[0067] (4) Schedule-related information: Schedule history information associated with user identification information is extracted from the user DB 12. If the user's schedule history indicated by the schedule history information does not include the user's schedule indicated by the schedule information received from the user terminal 70, schedule-related information regarding the user's schedule is searched for by internet search, etc. If new schedule-related information is extracted, it is acquired as new information.

[0068] In the prompt setting step S82, the conversation generation instruction unit 32 sets the type of prompt that indicates the direction of the conversation between the user and the AI ​​agent based on the setting conditions. In this embodiment, the prompt setting unit 36 ​​extracts conversation situation information associated with user identification information from the user DB 12, and sets either an information provision prompt in which the AI ​​agent provides a topic to the user, or an information gathering prompt in which the AI ​​agent asks a question to the user, as the type of prompt, based on the new information acquired by the new information acquisition unit 34, the current time information indicating the current time, and the conversation situation information associated with user identification information extracted from the user DB 12.

[0069] In this embodiment, for example, the prompt type is set as follows, as shown in Figure 5.

[0070] (1) Settings based on the existence of new information for the user (1-1) If new information for the user exists, set to the information provision prompt. (1-2) If no new information for the user exists, set to the information collection prompt.

[0071] In other words, if there is new information for the user, it is assumed that the user wants to be notified of the new information by the AI ​​agent, so the AI ​​agent is set to provide information to the user.

[0072] (2) Settings based on the current time of day (2-1) If the current time of day indicated by the current time of day information is from morning to noon, set it to the information provision prompt. (2-2) If the current time of day indicated by the current time of day information is from evening to night, set it to the information collection prompt.

[0073] In other words, if the current time is morning to noon, it is assumed that the user would rather have the AI ​​agent initiate a conversation, so the prompt is set to an information-providing prompt where the AI ​​agent offers topics to discuss with the user. On the other hand, if the current time is evening to night, it is assumed that the user would rather have the AI ​​agent listen to them, so the prompt is set to an information-gathering prompt where the AI ​​agent asks questions to the user.

[0074] (3) Settings based on the content of the conversation between the user and the AI ​​agent (3-1) If it is determined from the content of the conversation between the user and the AI ​​agent that it is an appropriate conversation situation for the AI ​​agent to provide a topic to the user, that is, if the conversation situation indicated by the conversation situation information is an information provision situation in which it is an appropriate conversation situation for the AI ​​agent to provide a topic to the user, then it will be set to an information provision prompt. (3-2) If it is determined from the content of the conversation between the user and the AI ​​agent that it is an appropriate conversation situation for the AI ​​agent to ask a question to the user, that is, if the conversation situation indicated by the conversation situation information is an information gathering situation in which it is an appropriate conversation situation for the AI ​​agent to ask a question to the user, then it will be set to an information gathering prompt.

[0075] In other words, if the conversation between the user and the AI ​​agent determines that it is an appropriate conversational situation for the AI ​​agent to provide a topic to the user, it can be assumed that the user wants the AI ​​agent to say something, and the prompt is set to an information provision prompt where the AI ​​agent provides a topic to the user. On the other hand, if the conversation between the user and the AI ​​agent determines that it is an appropriate conversational situation for the AI ​​agent to ask the user a question, it can be assumed that the user wants the AI ​​agent to listen to them, and the prompt is set to an information gathering prompt where the AI ​​agent asks the user a question.

[0076] Topic Candidate Provision Step S84: In the conversation generation instruction unit 32, the topic candidate provision unit 38 provides topic candidates corresponding to the type of prompt that indicates the direction of the conversation between the configured user and the AI ​​agent.

[0077] Topic Candidate Weighting Step S86 In the conversation generation instruction unit 32, the topic candidate weighting unit 42 performs weighting on the topic candidates provided by the topic candidate provision unit 38 in accordance with the type of prompt, based on the perspective of weighting.

[0078] In this embodiment, for information provision prompts that provide topics to the user from the AI ​​agent, topic candidates are provided as follows, for example as shown in Figure 6A, and the provided topic candidates are weighted based on the perspective of weighting.

[0079] Potential topics include the following: (α) Major categories (1) Topics related to others (2) Topics related to the user (β) Minor categories (1-1) Topics related to notifications from other applications on the user's device (1-2) News-related topics such as news and releases that match the user's preferences (1-3) Location-related topics related to the user's current location and surrounding area (1-4) Schedule-related topics related to the user's schedule (2-1) Topics related to the user's occupation (2-2) Topics related to the user's health status (γ) Minor categories Further subdivisions of the minor categories

[0080] The weighting criteria include the following, and the weighting methods based on each criterion are as follows.

[0081] (a-2) User's familiarity with the AI ​​agent If the user's familiarity with the AI ​​agent is relatively low and below a predetermined first threshold, it can be determined that the AI ​​agent has not been able to build a relationship with the user. In this case, even when providing topics of conversation, providing topics about the user themselves may give the user a rude impression. Therefore, for (α) major categories, (1) emphasis is placed on providing topics other than those about the user. If the familiarity exceeds the first threshold, it can be determined that the AI ​​agent has been able to build a certain relationship with the user. Therefore, for (α) major categories, (2) emphasis is placed on providing topics about the user themselves.

[0082] If the user's level of familiarity with the AI ​​agent exceeds the first threshold but is below the second threshold (moderate level), it can be determined that the AI ​​agent has not yet established a sufficiently close relationship with the user. In this case, providing topics that are relatively sensitive to the user may give the user an impression of rudeness. Therefore, for the (β) intermediate category, emphasis is placed on providing topics that are not relatively sensitive to the user, such as (2-1) topics related to the user's occupation. If the level of familiarity exceeds the second threshold, it can be determined that the AI ​​agent has established a sufficiently close relationship with the user. Therefore, emphasis is placed on providing topics that are relatively sensitive to the user, such as (2-2) topics related to the user's health.

[0083] (a-2) If new information exists for the user, it is assumed that the user wishes to be notified of the new information by the AI ​​agent. Therefore, (α) for major categories, emphasis will be placed on providing topics other than those concerning the user themselves, and (β) for minor categories, emphasis will be placed on providing topics from (1-1) to (1-4) above that contain new information for the user.

[0084] In this embodiment, for information gathering prompts in which the AI ​​agent asks questions to the user, topic candidates are provided as follows, for example, as shown in Figure 6B, and the provided topic candidates are weighted based on a weighting perspective.

[0085] Potential topics include the following: (α) Major categories (3) Questions about people other than the user (4) Questions about the user (β) Minor categories (3-1) Questions about the weather (4-1) Questions to check the user's current condition (4-2) Questions to check the user's current schedule (4-3) Questions to check the user's current situation (4-4) Questions about items for which the user's personal information has not been obtained (4-5) Detailed questions about items for which the user's personal information has been obtained (γ) Minor categories Further subdivisions of each minor category

[0086] The weighting criteria include the following, and the weighting methods based on each criterion are as follows.

[0087] (b-1) User's familiarity with the AI ​​agent If the user's familiarity with the AI ​​agent is below the second threshold of moderate level, it can be determined that the AI ​​agent has not yet established a sufficiently close relationship with the user, and asking questions about the user may give the user the impression that it is impolite. Therefore, for (α) major categories, (1) questions about people other than the user are given priority, and for (β) minor categories, innocuous questions, such as (3-1) questions about the weather as described above, are given priority. If the familiarity level exceeds the second threshold, it can be determined that the AI ​​agent has established a sufficiently close relationship with the user, and therefore, for (α) major categories, (1) questions about the user are given priority.

[0088] If the user's level of familiarity with the AI ​​agent exceeds the second threshold but is below the relatively high third threshold, it can be determined that the AI ​​agent has not yet established a sufficiently close relationship with the user. In this case, questions that are relatively sensitive to the user may cause discomfort, so the (β) subcategory will be biased towards innocuous questions that are not relatively sensitive to the user, such as (4-1) questions to check the user's current condition, (4-2) questions to check the user's current schedule, and (4-3) questions to check the user's current situation. If the level of familiarity exceeds the third threshold, it can be determined that the AI ​​agent has established a sufficiently close relationship with the user. In this case, the AI ​​agent will be biased towards questions that are relatively sensitive to the user, such as (4-4) questions about items for which the user's personal information has not been obtained, and (4-5) detailed questions about items for which the user's personal information has been obtained.

[0089] (b-2) The degree to which the user's personal information items are complete and the completeness of the content of each item If the degree to which the user's personal information items are complete is low, the AI ​​agent will broaden the topics of conversation with the user and build a broader relationship with the user. For (β) subcategories, for example, emphasis will be placed on questions related to items for which the user's personal information has not been obtained. If the degree to which the user's personal information items are complete is high, the AI ​​agent will deepen the topics of conversation with the user and build a deeper relationship with the user. For (β) subcategories, for example, emphasis will be placed on detailed questions related to items for which the user's personal information has been obtained.

[0090] (b-3) User's Positivity in Most Recent Conversations with the AI ​​Agent If the user's positivity in most recent conversations with the AI ​​agent is low, it can be determined that the user is not receptive to conversations with the AI ​​agent. If the AI ​​agent asks the user for details about their personal information, the user may become resentful. To change the topic, the (β) subcategory will be biased towards questions about items for which the user's personal information has not been obtained, such as (4-4). If the positivity is high, it can be determined that the user is receptive to conversations with the AI ​​agent. It can be assumed that the user is actually willing to talk to the AI ​​agent about details about their personal information. Therefore, the (β) subcategory will be biased towards questions about items for which the user's personal information has been obtained, such as (4-5).

[0091] In the prompt generation step S88, the conversation generation instruction unit 32 generates a speech generation prompt that instructs the conversation generation AI to generate speech text data indicating the content of the AI ​​agent's utterances to the user. As described above, this speech generation prompt ensures that the AI ​​agent's utterances are natural in light of the conversation history and match the characteristics of the AI ​​agent's character. It also instructs the prompt setting unit 36 ​​to select a topic from the topic candidates provided by the topic provision unit in accordance with the type of prompt, i.e., the direction of the conversation, and then sets a topic based on the weight assigned by the topic weighting unit, and then instructs the prompt setting unit 36 ​​to produce utterances corresponding to that topic.

[0092] Thus, the topics selected for utterances are weighted in accordance with the direction of the conversation, with a particular emphasis on the most suitable topics based on the user's familiarity with the AI ​​agent. This allows the AI ​​agent to build a close relationship with the user, and while some topics other than the optimal ones are selected with a certain degree of randomness within an appropriate range, the content of the AI ​​agent's utterances to the user is avoided to become repetitive, improving the user's motivation for conversation with the AI ​​agent and enabling the building of a closer relationship between the user and the AI ​​agent.

[0093] In the embodiment described above, the direction of the conversation is set based on the set conditions, and then each candidate topic is weighted based on the perspective of weighting corresponding to the direction of the conversation. In this way, the AI ​​agent generates utterances corresponding to the set topic from among the candidate topics according to the weights assigned to each candidate topic. As a result, even if the conversation with the AI ​​agent continues, it is possible to avoid the AI ​​agent repeating similar conversations and to improve the motivation to continue the conversation with the AI ​​agent.

[0094] In particular, when setting the direction of the conversation, the direction of the conversation is set based on the content of the conversation between the user and the AI ​​agent, so that the AI ​​agent can have an appropriate conversation with the user. Also, since the direction of the conversation is set to gather information by asking questions to the user based on the content of the conversation between the user and the AI ​​agent, questions to the user can be asked in an appropriate conversational situation.

[0095] Furthermore, in setting topics, as the user's familiarity with the AI ​​agent increases, the weight of topics related to the user, especially topics related to the user's personal information, is increased. This prevents situations where the conversation about the user is rude or offensive if a sufficiently close relationship has not yet been established between the user and the AI ​​agent. Additionally, as the level of positivity in the user's recent conversations with the AI ​​agent increases, the weight of topics related to the user's personal information is increased. This prevents situations where the conversation about the user's personal information is frowned upon if the user is not receptive to the conversation, thus avoiding causing resentment.

[0096] In this way, it becomes possible to build a close relationship between the user and the AI ​​agent, further improving the user's motivation to converse with the AI ​​agent.

[0097] In the embodiment described above, a speech generation prompt is generated and input to a conversation generation AI, and after receiving output of speech text data from the conversation generation AI, the speech text data is converted into speech audio information indicating the AI ​​agent's speech based on the AI ​​agent's voice information. Alternatively, after converting conversation audio information indicating the user's conversation voice into conversation audio text data, a response generation prompt including the conversation audio text data is generated and input to the conversation generation AI, and after receiving output of response text data from the conversation generation AI, the response text data is converted into response audio information indicating the AI ​​agent's response based on the AI ​​agent's voice information. Alternatively, a response generation prompt including conversational audio information representing the user's speech and audio information of the AI ​​agent may be generated and input to a conversation generation AI, and response audio information representing the AI ​​agent's response may be output from the conversation generation AI. In this case, the utterance audio information, conversational audio information, and response audio information may be stored as conversation history information.

[0098] The distinctive disclosures of this application can be summarized as follows: The first disclosure is a conversational device comprising a conversational unit that instructs a task processing device, which processes tasks using artificial intelligence in response to task processing instructions, to generate conversational content for a virtual agent to a user, and acquires the conversational content generated by the task processing device and outputs it to a user terminal device, wherein the conversational unit instructs the task processing device to generate conversational content corresponding to a topic set from among the topic candidates, based on the weights assigned to each topic candidate from the perspective of weighting.

[0099] In this disclosure, the virtual agent generates conversation content corresponding to a selected topic from among the topic candidates, based on weights assigned to each topic candidate from a weighting perspective. As a result, even if the conversation with the virtual agent continues, it is possible to avoid the virtual agent repeating similar conversations and improve the motivation to continue talking with the virtual agent.

[0100] The second disclosure is a conversation provider according to the first disclosure, wherein the weighting aspect includes the user's familiarity with the virtual agent, the topics include topics about the user themselves, and the weight of candidate topics about the user themselves increases as the familiarity increases.

[0101] This disclosure states that as the user's familiarity with the virtual agent increases, the weight of potential topics related to the user themselves is increased. This prevents situations where the user is subjected to conversation about themselves before a certain level of familiarity has been established between the user and the virtual agent, potentially giving the user an impression of rudeness. It also makes it possible to build a closer relationship between the user and the virtual agent and further enhance the motivation for conversation with the virtual agent.

[0102] The third disclosure is a conversation provider as described in the second disclosure, in which topics concerning the user themselves include topics concerning the user's personal information, and the weight of candidate topics concerning the user's personal information increases as the level of intimacy increases.

[0103] This disclosure increases the weight of potential topics related to the user's personal information as the user's intimacy with the virtual agent increases. This avoids situations where the user's personal information is discussed and causes discomfort to the user before a sufficiently close relationship has been established between the user and the virtual agent. It also makes it possible to build a closer relationship between the user and the virtual agent and further improve the motivation for conversation with the virtual agent.

[0104] The fourth disclosure is a conversation provider according to the third disclosure, wherein the weighting aspect includes the degree of positivity of the user's most recent conversation with the virtual agent, and topics relating to the user's own personal information include topics relating to details of the user's own personal information, and the weight of candidate topics relating to details of the user's own personal information increases as the degree of positivity increases.

[0105] This disclosure increases the weight of potential topics related to the user's personal information as the positive impact of the user's recent conversations with the virtual agent increases. This avoids situations where the user is reluctant to engage in conversation and is subjected to discussions about their personal information, thereby preventing them from becoming averse to the conversation. It also makes it possible to build a sufficiently close relationship between the user and the virtual agent and further enhance the user's motivation to converse with the virtual agent.

[0106] The fifth disclosure is that the conversation unit is a conversation-providing device as described in the first disclosure, which sets a direction of conversation from among a plurality of directions of conversation based on set conditions, and weights each candidate topic based on the weighting perspective corresponding to the direction of conversation.

[0107] In this disclosure, the direction of conversation is set from among multiple directions of conversation based on predefined conditions, and then each candidate topic is weighted based on a weighting perspective corresponding to the direction of conversation. This further prevents the virtual agent from repeating similar conversations and further improves the motivation to converse with the virtual agent.

[0108] The sixth disclosure is a conversation provider according to the fifth disclosure, wherein the setting conditions include setting the direction of conversation based on the content of the conversation between the user and the virtual agent from among the plurality of conversation directions.

[0109] In this disclosure, the direction of the conversation is set based on the content of the conversation between the user and the virtual agent, from among several possible conversation directions. This allows the virtual agent to engage in appropriate conversations with the user and further improves the user's motivation to converse with the virtual agent.

[0110] The seventh disclosure is that the conversational direction is a conversational device as described in the sixth disclosure, which includes gathering information by asking questions to the user.

[0111] This disclosure sets the direction of the conversation to information gathering that involves asking questions to the user based on the content of the conversation between the user and the virtual agent. This allows questions to be asked to the user in an appropriate conversational context, fostering a close relationship between the user and the virtual agent and further improving the user's motivation to converse with the virtual agent.

[0112] 10...Conversation provision server 12...User DB 14...Agent DB 16...Agent designation unit 18...Conversation analysis instruction unit 22...Conversation status setting unit 24...Intimacy setting unit 26...Positivity setting unit 28...Personal information setting unit 32...Conversation generation instruction unit 34...New information acquisition unit 36...Prompt setting unit 38...Topic candidate provision unit 42...Topic candidate weighting unit 44...Prompt generation unit 46...Voice generation unit 48...Voice analysis unit 60...Generation server 62...Conversation analysis unit 64...Conversation generation unit 70...User terminal 72...Application launch unit 74...Agent selection unit 76...Image generation unit 78...Image display unit 82...Related information collection unit 84...Conversation execution unit 86...Voice output unit 88...Voice acquisition unit

Claims

1. A conversational device comprising a task processing device that processes tasks using artificial intelligence in response to task processing instructions, a conversational unit that instructs the task processing device to generate conversational content for a virtual agent to a user, and acquires the conversational content generated by the task processing device and outputs it to a user terminal device, wherein the conversational unit instructs the task processing device to generate conversational content corresponding to a set topic from among the topic candidates, based on the weights assigned to each topic candidate from the perspective of weighting.

2. The conversation provider according to claim 1, wherein the weighting considerations include the user's familiarity with the virtual agent, the topics include topics concerning the user themselves, and the weight of candidate topics concerning the user themselves is increased as the familiarity increases.

3. The conversation-providing device according to claim 2, wherein the topics relating to the user include topics relating to the user's personal information, and the weight of candidate topics relating to the user's personal information increases as the level of intimacy increases.

4. The conversation-providing device according to claim 3, wherein the weighting criteria include the degree of positivity of the user's most recent conversation with the virtual agent, topics relating to the user's own personal information include topics relating to details of the user's own personal information, and the weight of candidate topics relating to details of the user's own personal information increases as the degree of positivity increases.

5. The conversation unit sets a direction of conversation from among a plurality of directions of conversation based on set conditions, and assigns weights to each candidate topic based on the weighting perspective corresponding to the direction of conversation, as described in claim 1.

6. The conversation provision device according to claim 5, wherein the setting conditions include setting the direction of conversation from among the plurality of conversation directions based on the content of the conversation between the user and the virtual agent.

7. The conversation-providing device according to claim 6, wherein the direction of the conversation includes gathering information by asking questions to the user.

Images