Chatbot program and chatbot device

Abstract

We provide users with a sufficient amount of information that is easy for them to understand, tailored to their product knowledge and experience. [Solution] The chatbot device 100 acquires user-specific information related to the vehicle using the user vehicle-related information acquisition unit 112, receives vehicle-related queries from the user terminal using the user query input unit 113, and based on the received queries, the prompt generation unit 115 generates prompts for the large-scale language models 200 and 300. The original response, which is the answer to the prompt, is received by the AI ​​response information acquisition unit 117. Based on the user-specific information, the response information generation units 115, 116, 117, and 118 generate response information by changing the degree to which at least one of the vocabulary range, amount of information, and type of information decreases or increases, corresponding to at least one of the user's amount of knowledge related to the vehicle, the user's vehicle ownership history, vehicle usage status, and vehicle ownership status.

Description

Technical Field

[0005] , ,

[0001] The present disclosure relates to a chatbot program and a chatbot device.

Background Art

[0002] In information processing tasks where complexity and diversity are increasing, the usefulness of large language models (LLMs) is being recognized. In recent years, systems that support the sale of products to customers using mobile terminals such as smartphones and tablet terminals are known.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] By the way, when a customer wishes to obtain information on various products, if a uniform response is given to the questions sent by the customer, depending on the customer's product knowledge and experience, the content of the response may be felt to be complex and satisfaction may not be obtained, or the response may be felt to be redundant.

Means for Solving the Problems

[0005] A chatbot program according to one aspect of the present invention is executed by a computer comprising a processor and a memory unit, and involves the computer performing the following actions: acquiring user-specific information, which is information unique to the user related to a vehicle, from the user-specific information memory unit; receiving vehicle-related queries entered by the user via the user terminal from a user terminal operated by the user; and generating response information, wherein, based on the query, the program generates a prompt, which is an input sentence to a large-scale language model, and sends it to the large-scale language model; receiving the original response to the prompt from the large-scale language model; and generating response information to send to the user terminal based on the original response, wherein, based on the user-specific information, the program generates response information that decreases or increases, to a certain extent, the range of vocabulary, the amount of information, and the type of information constituting the response information, corresponding to at least one of the user's knowledge amount related to the vehicle, the user's vehicle ownership history, the user's vehicle usage status, and the user's vehicle ownership status; and sending the response information to the user terminal.

[0006] Another aspect of the present invention is a chatbot device comprising: a user vehicle-related information acquisition unit that acquires user-specific information, which is information unique to the user related to a vehicle, from a user-specific information storage unit; a user query input unit that receives vehicle-related queries entered by the user via a user terminal operated by the user; a response information generation unit that, based on the query, generates a prompt which is an input sentence to a large-scale language model and sends it to the large-scale language model, receives an original response to the prompt from the large-scale language model, and generates response information to be sent to the user terminal based on the original response, while decreasing or increasing at least one of the range of vocabulary, amount of information, and type of information that constitute the response information, based on the user-specific information, corresponding to at least one of the amount of knowledge the user has about vehicles, the user's vehicle ownership history, the user's vehicle usage status, and the user's vehicle ownership status; and a response information output unit that sends the response information to the user. [Effects of the Invention]

[0007] According to this disclosure, the chatbot system can provide answers tailored to individual users. [Brief explanation of the drawing]

[0008] [Figure 1] A conceptual diagram illustrating the general configuration of a chatbot system. [Figure 2] A conceptual diagram that provides a general overview of an example data structure for a customer database. [Figure 3] A conceptual diagram illustrating an example of the data structure of user data stored in a customer database. [Figure 4] A block diagram illustrating the functional blocks of a server and the components that connect to the server via a network. [Figure 5] A flowchart illustrating the general processing steps of chatbot response processing, which is performed by the CPU of the server providing the chatbot service to the user. [Modes for carrying out the invention]

[0009] Embodiments of this disclosure will be described below with reference to Figures 1 to 5. In all the figures described below, common components are denoted by the same reference numerals, and repeated descriptions are omitted. Not all components shown in the following embodiments are necessarily essential components of this disclosure.

[0010] Figure 1 is a conceptual diagram illustrating how the server 100 constituting the chatbot system is connected via a network (NW) to an external LLM 200, an in-house Generative AI system 300, a customer database 400, and a user terminal 500. In this embodiment, the chatbot system is a system that responds to user questions mainly about vehicles such as passenger cars manufactured and sold by the company, and services related to vehicles.

[0011] The server 100 comprises a CPU 101, memory 102, I / O 104, storage 105, and a data bus 103 that enables data exchange between these elements.

[0012] The chatbot program for providing chatbot services to users may, for example, be stored in storage 105, read from storage 105 when server 100 starts up, loaded into memory 102, and executed by CPU 101. Alternatively, memory 102 may include ROM or flash memory, and the chatbot program stored in this ROM or flash memory may be loaded into memory 102 and executed by CPU 101. Furthermore, a program obtained by reading from an optical disc drive or card reader (not shown) or by downloading via a network NW may be loaded into memory 102 and executed by CPU 101.

[0013] Furthermore, by distributing all or part of each hardware configuration across multiple computers and connecting them to each other via a network (NW), a server 100 can be virtually realized. In other words, the concept of server 100 includes not only systems housed in a single chassis or case, but also virtualized computer systems.

[0014] External LLM200 refers to LLMs provided by third parties, for example. An LLM (Large Language Model) is a large-scale artificial intelligence model used in the field of natural language processing (NLP), built using a large set of text data and deep learning techniques. By learning from large amounts of text data (web pages, books, articles, etc.), LLMs can understand patterns in human language and effectively perform natural language generation (NLG) tasks.

[0015] The following explanation uses LLM as an external source when providing a chatbot service to users as an example, but other AI-powered external services such as ChatGPT, Google Gemini®, Cerence®, and Claude3® may also be used.

[0016] The in-house AI generation system 300 is a system built by accumulating the company's own information as a search source 310 for the chatbot service. It contains a wealth of information on the products and services the company sells and provides, as well as related operation manuals and other documents, and past FAQs (Frequently Asked Questions). This makes it possible to improve the accuracy of AI-generated answers and reduce hallucination. In addition to outputting general answers to questions from users driving vehicles, it can also provide more specific responses, such as, "Of the multiple steering switches on the right steering spoke, please press the switch with the green icon printed on it, located on the lower left." Furthermore, because it is an in-house system, it has the advantage of being able to suppress cost increases even if the number of transactions when interacting with the AI ​​during a chatbot session increases.

[0017] The customer database 400 is a database that stores login information such as passwords, linked to the user ID used when a user logs into the chatbot service, as well as user data which will be explained in detail later.

[0018] The user terminal 500 is an information processing device operated by a user utilizing the chatbot service provided by the server 100. The user terminal 500 can be, for example, a smartphone, tablet, PDA (Personal Digital Assistant), or personal computer, and it is also possible to use an in-vehicle communication module.

[0019] The network NW for transmitting information between the elements described above can utilize the Internet. Note that the in-house generated AI system 300, the customer database 400, and the server 100 may be connected via a closed network service without using the Internet, or may be connected using VPN technology. Alternatively, the in-house generated AI system 300 and the customer database 400 may be included inside the server 100. For the customer database 400, it is desirable to use strong encryption technology to protect it in order to enhance information security.

[0020] FIG. 2 is a conceptual diagram schematically explaining the data structure of the customer database 400. The customer database 400 is configured to include a user ID 402, login information 404, user data 406, and the like. The login information 404 and the user data 406 are stored in association with the user ID 402. The user ID 402 stores data that is referred to when collating the user ID transmitted from the user terminal 500 at the start of a chatbot session described later. The login information 404 stores information such as passwords corresponding to individual user IDs 402 and past login dates and times.

[0021] FIG. 3 is a conceptual diagram showing an example of the information held as the user data 406. The user data 406 stores information that can identify the amount of knowledge related to the user's vehicle, the user's vehicle ownership history, the user's vehicle usage status, and the user's vehicle ownership status. The user data 406 further stores information regarding questions issued by the user and answers thereto in past chatbot sessions, and session information regarding each chatbot session.

[0022] The following describes examples of information recorded as user data 406, referring to Figure 3. Vehicle ownership history information includes information about the types of vehicles the user has previously owned, the period of ownership, and the specifications of those vehicles. Information about the vehicle currently in use includes information about the drive system, body shape, engine type, passenger capacity, usage patterns, and ADAS (Advanced Driver-Assistance Systems) functions. Supplementary information includes information about the user's residence, chatbot usage history, and knowledge level regarding vehicles.

[0023] The drivetrain information can include details that identify the vehicle's configuration, such as F / F (front-engine / front-wheel drive), F / R (front-engine / rear-wheel drive), M / R (mid-engine / rear-wheel drive), R / R (rear-engine / rear-wheel drive), and AWD (all-wheel drive).

[0024] Information about vehicle body shape can include categories such as sedans (box-shaped), vehicles with a canvas roof, hatchbacks, station wagons (including minivans and one-box cars), etc. This information about vehicle body shape can be further categorized and stored as SUVs, minivans, one-box cars, deck vans, off-road vehicles, kei cars, kei wagons, kei vans, etc.

[0025] The information regarding the prime mover may include identifying details such as gasoline engines, diesel engines, hydrogen engines that use hydrogen to generate power, CNG engines that use CNG (compressed natural gas) to generate power, HEVs (hybrid electric vehicles powered by an internal combustion engine and an electric motor), PHEVs (plug-in hybrid electric vehicles), BEVs (battery electric vehicles), and FCEVs (fuel cell electric vehicles).

[0026] The passenger capacity information can include numerical data such as the passenger capacity of the vehicle currently in use by the user, for example, 2, 4, 5, 6, 7, 8, or 10. Alternatively, it can include information that specifies the number of seats that can be occupied in the front / rear seats, or the number of seats that can be occupied in the front / second / third rows, such as 2 / 3, 2 / 2 / 3, etc.

[0027] Information regarding usage patterns can include details that identify the purpose and manner in which the user most frequently uses their current vehicle, such as short-distance trips, commuting, leisure, and long-distance drives. ADAS information can include information that identifies what driver assistance and safety features the vehicle currently being used by the user is equipped with. For example, it can include information that identifies functions such as ACC (Adaptive Cruise Control: a vehicle speed control system that automatically accelerates and decelerates while maintaining a safe distance from the vehicle in front), LDW (Lane Departure Warning: a lane departure warning system), LKAS (Lane Keep Assist System: a lane keeping assist system), RCTA (Rear Cross Traffic Alert: a system that detects vehicles crossing behind the vehicle when reversing and alerts the driver), CMBS (Collision Mitigation Braking System: a collision mitigation braking system), and PPS (Pedestrian Protection System: a vehicle pedestrian protection system).

[0028] Regarding information about the user's residence, the user's residence is, • In apartment buildings, whether or not there is a parking lot, and whether or not there is a charging facility for BEVs. • The information can include details that identify whether a detached house has a parking space or not, whether it has a charging facility for BEVs or not, and whether it has V2H (Vehicle to Home: a system that allows the use of electricity stored in the batteries of BEVs and PHEVs in the home) or not.

[0029] Information regarding chatbot usage history can include questions, answers, and session information. Question information may include details that identify the content of the questions asked by the user when the chatbot session was opened, and the level of technical jargon (vocabulary) used. Answer information may include details that identify the vocabulary, amount of information, and types of information included in the answer. Session information may include details that identify the user's evaluation and level of understanding of the chatbot's answers, as well as when the chatbot session started and ended.

[0030] Knowledge level information is used to determine the level of technical jargon (vocabulary), complexity of explanation, and length of explanation that the chatbot will use in its responses related to vehicles. This information is referenced to ensure that the explanation is easy for the user to understand and is not redundant.

[0031] A user's level of knowledge changes depending on the length of time they have owned a vehicle and their experience. Therefore, it is desirable that information regarding the level of knowledge be configured to be up-to-date at any time. For example, information regarding a user's level of knowledge about their vehicle can be updated by referring to session information accumulated as the user repeatedly engages in chatbot sessions. For example, during or at the end of a chatbot session, the user may provide feedback such as, "I would like a simpler explanation with fewer technical terms," ​​or "I would like a more detailed explanation with more technical terms," ​​and session information may be generated according to that feedback. Consequently, information regarding the level of knowledge can be updated.

[0032] Figure 4 is a block diagram illustrating the chatbot processing unit implemented by the CPU 101 of the server 100 executing the chatbot program, and shows it together with external elements connected to the server 100 via the network NW.

[0033] The user identification unit 111 exchanges a user ID and password with the user terminal 500 that is attempting to connect to the server 100 and start a chatbot session, and identifies and authenticates the user based on the user ID 402 and login information 404 obtained by accessing the customer database 400. The user vehicle-related information acquisition unit 112 acquires user data 406 corresponding to the user identified by the user identification unit 111 from the customer database 400.

[0034] If user data 406 is not registered in the customer database 400, the user vehicle-related information acquisition unit 112 outputs a chat message to the user operating the logged-in user terminal 500 (hereinafter referred to as the chat user) requesting them to input the necessary information. The chat user responds to the chatbot's questions by inputting user vehicle-related information that will serve as the basis for registering user data 406, as explained with reference to Figure 3, in the customer database 400.

[0035] The user query input unit 113 receives queries entered by chat users. Specifically, the user query input unit 113 receives queries entered by chat users via the user terminal 500 from the user terminal 500. The user query input unit 113 outputs the queries received from the user terminal 500 to the prompt generation unit 115. Queries entered by chat users may be text-based or voice-based. The prompt generation unit 115 analyzes the queries entered by chat users (received from the user terminal 500) and generates prompts to be output to the external LLM 200 or the in-house AI generation system 300. In this embodiment, an example is described in which there are multiple external LLM 200s and the in-house AI generation system 300, but it is also possible to have only one of them as the destination for outputting prompts.

[0036] If there are multiple candidates for the destination to output a prompt, the server 100 may decide whether to output the prompt to the external LLM200 or the in-house AI generation system 300, depending on the query from the chat user and the user vehicle-related information specific to the chat user, and the prompt generation unit 115 may generate an input sentence corresponding to that output destination.

[0037] The prompt output unit 116 outputs the input sentence, i.e., the prompt, generated by the prompt generation unit 115 to one of the output destinations determined by the server 100. The AI ​​response information acquisition unit 117 receives the response from the external LLM 200 or the in-house AI generation system 300 to which the prompt output unit 116 has output a prompt. In this specification, the response received by the AI ​​response information acquisition unit 117 from the external LLM 200 or the in-house AI generation system 300 to which the prompt output unit 116 has output a prompt is referred to as the original response.

[0038] The response information editing unit 118 edits the original response acquired by the AI ​​response information acquisition unit 117 based on the chat user-specific user data 406 acquired by the user vehicle-related information acquisition unit 112, and generates response information that is to be presented to the chat user.

[0039] More specifically, the response information editing unit 118 performs the editing process as follows: Based on the knowledge amount information explained with reference to Figure 3 and the chatbot usage history, it determines the amount of knowledge related to the chat user's vehicle, and then determines the level of detail to provide, the extent to which technical terms should be included (vocabulary range), and the type of information to provide. Alternatively, or in addition to this, it is also possible to determine the amount of knowledge related to the chat user's vehicle by referring to the user's vehicle ownership history information, and then determine the level of detail to provide, the extent to which technical terms should be included (vocabulary range), and the type of information to provide.

[0040] For example, suppose a chat user asks a question not using technical jargon like "I want to know how to use the ACC function," but rather "I want to know how to use the function that drives at a constant speed and then slows down to follow the vehicle in front when the distance to the vehicle in front gets close." In this case, if the chat user's level of knowledge is considered to be that of a beginner, and the question does not use the technical term ACC, then it can be determined that the chat user is not familiar with technical terms, and the original answer can be edited by replacing the technical term ACC with simpler language.

[0041] Furthermore, by referring to the drive system, body shape, engine, passenger capacity, usage pattern, and ADAS information of the vehicle currently used by the chat user, it is possible to determine the scope of the vocabulary and the types of information to provide. For example, suppose a chat user asks, "I want to know how to use LKAS." If, after referring to the chat user's unique user data 406, it is found that the chat user's vehicle is equipped with LKAS functionality, the response information editing unit 118 can omit the explanation of LKAS functionality in the original response and edit the response information to focus on the specific operation method for setting / deactivating LKAS in the chat user's vehicle.

[0042] Furthermore, suppose the chat user's question was, "I want to know about BEV charging facilities." Then, referring to user data 406 specific to the chat user, it is determined, based on the chat user's residence information, that they live in an apartment building with a parking space and do not have a charger installed. The response information editing department 118 may edit the response information by reducing or omitting information about home charging facilities in the original response, and narrowing the scope of vocabulary, amount of information, and type of information to at least one of these.

[0043] By doing so, based on user data 406, it becomes possible to generate response information in which the range of vocabulary, the amount of information, and the types of information constituting the response information are reduced or increased to a certain degree, corresponding to at least one of the user's knowledge related to vehicles, the user's vehicle ownership history, the user's vehicle usage status, and the user's vehicle ownership status.

[0044] The response information output unit 119 outputs the response information edited by the response information editing unit 118 to the user terminal 500. When a chatbot session ends, the session information extraction unit 120 extracts information that can identify the queries and answers exchanged between the chatbot and the user terminal 500 during the chatbot session, information related to the chat user's evaluation of the chatbot's answers, information that can identify the chat user's level of understanding, and information that can identify when the chatbot session started and ended, and appends this information to the user data 406 in the customer database 400.

[0045] In this case, the session information extraction unit 120 may update the information regarding the amount of knowledge in the user data 406 if it determines that the amount of knowledge of the chat user has increased as the number of times the chat user uses the chatbot session increases. Alternatively, the information regarding the amount of knowledge may be updated in response to feedback from the chat user during the chatbot session, such as a request to "use more technical terms and make the explanation more concise."

[0046] Figure 5 is a flowchart illustrating the flow of chatbot response processing executed by the CPU 101 of server 100. When a chat user operates the user terminal 500 and a chatbot session is started, the CPU 101 obtains the user ID from the user terminal 500 in S500.

[0047] In S502, CPU 101 accesses customer database 400 to obtain login information 404 and user data 406 corresponding to the chat user's login ID. It then compares the password information contained in login information 404 with the password entered by the chat user to perform authentication.

[0048] In S504, the CPU 101 determines whether or not user data 406 specific to the chat user exists in the customer database 400. If it is determined in S504 that user data 406 does not exist, the process proceeds to S506. In S506, the CPU 101 processes the user data input. In this user data input process, the CPU 101 outputs a chat message to the user terminal 500 prompting the chat user to input the information that forms the basis of the user data 406. Alternatively, a separate screen may be opened to prompt the chat user for information input. For chat users who do not have time to input an answer to a question from the chatbot and who urgently need information from the chatbot, the chat user may be asked to input at least the amount of vehicle-related knowledge they acknowledge.

[0049] On the other hand, if S504 determines that user data exists, the process proceeds to S508. In S508, CPU 101 receives the query issued by the chat user. Specifically, CPU 101 receives the query entered by the chat user via user terminal 500 from user terminal 500. In S510, CPU 101 analyzes the content of the query from the chat user that was entered in S508 (received from user terminal 500), determines the LLM to which the prompt will be output, and generates a prompt to be output to that LLM. If there is only one LLM connected to server 100, the process of determining the output destination LLM can be omitted.

[0050] The method for determining which LLM to output the prompt to can be based on each LLM's area of ​​expertise, the required level of detail in the response, etc. For example, suppose the query concerns a vehicle owned by the chat user, and that the user's vehicle is a product of the company providing the chatbot service. In such a case, the in-house AI system 300 could be chosen as the prompt output destination. Doing so would allow for the provision of more accurate information to the chat user.

[0051] In S512, CPU101 outputs a prompt to the LLM that was determined as the output destination in S510, and in S514, it obtains the response output from the LLM in response to that prompt. In this specification, the response obtained from the LLM that output the prompt (including the in-house AI system) is referred to as the original response.

[0052] In S516, the CPU 101 processes the original response obtained in S514 based on the user data obtained in S502 or S506 to generate response information to send to the chat user. That is, it executes the function of the response information editing unit 118 as explained with reference to Figure 4. This makes it possible to generate response information that, based on the user data 406, decreases or increases the range of vocabulary, the amount of information, and the type of information that constitute the response information, while changing the degree of at least one of them, corresponding to at least one of the user's knowledge amount related to vehicles, the user's vehicle ownership history, the user's vehicle usage status, and the user's vehicle ownership status. In S518, CPU101 sends the response information generated in S516 to the user terminal 500 owned by the chat user.

[0053] If, in S520, it is determined that the chatbot session will continue, meaning that a further query has been entered by the chat user, the process returns to S508 and the series of processes described above from S508 to S520 continues. On the other hand, if it is determined in S520 that the chatbot session has ended, the CPU 101 records the chatbot usage history information in S522. The chatbot usage history information includes the information extracted by the session information extraction unit 120 as explained with reference to Figure 4. After completing the processing in S522, the CPU 101 enters a waiting state until a new chatbot session is started.

[0054] The above describes an example in which the original response output from the external LLM200 or the in-house AI generation system 300 is edited based on the chat user's user data 406. In this regard, it is also possible for the CPU 101 to refer to the user data 406 when generating a prompt to output to the external LLM200 or the in-house AI generation system 300 based on a query received from the user terminal 500 operated by the chat user, and to generate a prompt that makes it easier to obtain an original response corresponding to this user data 406. By doing so, it is possible to obtain response information that is reduced or increased in degree to at least one of the following: the range of vocabulary, the amount of information, and the type of information that constitute the response information, corresponding to at least one of the user's knowledge amount related to vehicles, the user's vehicle ownership history, the user's vehicle usage status, and the user's vehicle ownership status, based on the user data 406.

[0055] Furthermore, while the above describes an example where a chat user operating a user terminal 500 directly interacts with a server 100 providing a chatbot service, an operator may also be involved. For example, the operator may receive questions from the chat user and output prompts to an external LLM 200 or an in-house AI generation system 300. In this case, the response results from the external LLM 200 or the in-house AI generation system may be presented to the operator via the response information editing department 118. The operator can then refer to the presented response results and reply to the customer.

[0056] This embodiment can provide the following effects and advantages. (1) The chatbot program is to be executed by a computer 100 (Figure 1) which has a processor 101 and a memory unit 102, and is to acquire user-specific information 406 (Figures 2 and 3), which is user-specific information related to the vehicle, from a user-specific information memory unit 400 (Figures 1 and 4), receive vehicle-related queries entered by the user via a user terminal 500 (Figure 1) operated by the user (S508 in Figure 5), and generate response information (S510, S512, S514, S516 in Figure 5), and based on the query, generate prompts which are input sentences for the large-scale language models 200 and 300. The system sends the response to the large-scale language models 200 and 300, receives the original response to the prompt from the large-scale language models 200 and 300, and generates response information to send to the user terminal based on the original response. In doing so, the system generates response information that decreases or increases, to a certain extent, the range of vocabulary, the amount of information, and the type of information that constitute the response information, corresponding to at least one of the user's knowledge amount related to the user's vehicle, the user's vehicle ownership history, the user's vehicle usage status, and the user's vehicle ownership status, based on user-specific information 406 (Figure 3), and transmits the response information to the user terminal 500.

[0057] According to this, responses that are easy for the user to understand and less likely to be perceived as redundant can be sent to the user's terminal, depending on the user's unique background, such as the amount of knowledge the user has about vehicles, the user's vehicle ownership history, the user's vehicle usage, and the user's vehicle ownership status.

[0058] (2) If the chatbot program does not have user-specific information 406 in the user-specific information storage unit 400, it further causes the computer 100 to output a message to the user prompting the user to input user-specific information (S506 in Figure 5). According to this, even if user-specific information 406 does not exist in the user-specific information storage unit, an accurate response can be sent to the user.

[0059] (3) In the chatbot program, the generation of response information (S516 in Figure 5) is based on user-specific information 406, and the longer the user has owned a vehicle, the more information the response information will contain, by reducing the extent to which at least one of the vocabulary, amount of information, and types of information is reduced. According to this, it is possible to send accurate answers to users who are expected to have owned vehicles for a long time and possess a corresponding amount of knowledge.

[0060] (4) In the chatbot program, when generating response information (S516 in Figure 5), if the user has no history of owning an electric vehicle, the response information will be reduced by increasing the degree to which at least one of the vocabulary, amount of information, and types of information related to electric vehicles is reduced. According to this, it becomes possible to send users who have never owned an electric vehicle answers that are easier for them to understand, by reducing information that contains unfamiliar vocabulary or is difficult to understand.

[0061] (5) In the chatbot program, generating response information (S510, S512, S514, S516 in Figure 5) involves extracting information about the user's residence based on user-specific information 406 when the query includes matters related to charging electric vehicles, and then changing the extent to which at least one of the vocabulary range, amount of information, and type of information is reduced or increased based on the information about the user's residence to provide response information suitable for the user's residence. This makes it possible to provide information that is appropriate to the user's living situation.

[0062] (6) The chatbot device 100 includes a user vehicle-related information acquisition unit 112 (Figure 4) that acquires user-specific information 406 (Figures 2 and 3), which is user-specific information related to the vehicle, from a user-specific information storage unit 400 (Figures 1 and 4), a user query input unit 113 that receives vehicle-related queries entered by the user via the user terminal from the user terminal operated by the user, and a response information generation unit (115, 116, 117, 118 in Figure 4) that generates prompts, which are input sentences for the large-scale language models 200 and 300, based on the queries and sends them to the large-scale language models 200 and 300, and responds to the prompts. The system includes response information generation units 115, 116, 117, and 118 that receive the original response from large-scale language models 200 and 300, and generate response information to be sent to the user terminal 500 based on the original response. These units decrease or increase the range of vocabulary, the amount of information, and the type of information that constitute the response information, based on user-specific information 406 (Figure 3), corresponding to at least one of the user's knowledge amount related to vehicles, the user's vehicle ownership history, the user's vehicle usage status, and the user's vehicle ownership status, while changing the degree of each. The system also includes a response information output unit 119 that sends the response information to the user terminal.

[0063] According to this, it is possible to provide a chatbot device that can send answers to the user's terminal that are easy for the user to understand and do not feel redundant, according to the user's unique background, such as the amount of knowledge the user has about vehicles, the user's vehicle ownership history, the user's vehicle usage, and the user's vehicle ownership status.

[0064] The above description is merely an example, and the present invention is not limited by the embodiments and modifications described above, as long as the features of the present invention are not impaired. It is also possible to arbitrarily combine one or more of the above embodiments and modifications, and to combine modifications with each other. [Explanation of Symbols]

[0065] 100 Server, 101 CPU, 102 Memory, 103 Data Bus, 104 I / O, 105 Storage, 111 User Identification Unit, 112 User Vehicle-Related Information Acquisition Unit, 113 User Query Input Unit, 115 Prompt Generation Unit, 116 Prompt Output Unit, 117 AI Response Information Acquisition Unit, 118 Response Information Editing Unit, 119 Response Information Output Unit, 120 Session Information Extraction Unit, 200 External LLM, 300 In-house Generated AI System, 310 Search Source, 400 Customer Database, 402 User ID, 404 Login Information, 406 User Data, 500 User Terminal, NW Network

Claims

1. A chatbot program to be executed on a computer having a processor and a memory unit, This involves obtaining user-specific information, which is user-specific information related to the vehicle, from the user-specific information storage unit, The system receives vehicle-related queries entered by the user via the user terminal operated by the user, The method for generating response information involves generating a prompt, which is an input sentence to a large-scale language model, based on the query, and sending it to the large-scale language model; receiving an original response to the prompt from the large-scale language model; and generating the response information to be sent to the user terminal based on the original response, wherein, based on the user-specific information, the method involves decreasing or increasing, to a certain extent, the range of vocabulary, the amount of information, and the type of information that constitute the response information, corresponding to at least one of the user's knowledge amount related to vehicles, the user's vehicle ownership history, the user's vehicle usage status, and the user's vehicle ownership status. To transmit the aforementioned response information to the user terminal. A chatbot program characterized by causing the computer to execute the following.

2. A chatbot program according to claim 1, characterized in that, if the user-specific information is not present in the user-specific information storage unit, the computer further causes the computer to output a message to the user prompting the user to input the user-specific information.

3. The chatbot program according to claim 2, wherein the generation of the response information is characterized in that, based on the user-specific information, the longer the user's vehicle ownership period, the more the response information contains more information by reducing the degree to which at least one of the vocabulary, the amount of information, and the types of information is reduced.

4. The chatbot program according to claim 2, wherein generating the response information is characterized in that, based on the user-specific information, if the user has no history of owning an electric vehicle, the amount of information and the type of information related to the electric vehicle are reduced to an increased degree to which the response information is further reduced.

5. In the chatbot program according to claim 2, generating the response information involves, when the query includes matters relating to the charging of an electric vehicle, extracting information relating to the user's residence based on the user-specific information, A chatbot program characterized by changing the extent to which at least one of the vocabulary range, the amount of information, and the type of information is reduced or increased based on information about the user's residence, thereby providing response information suitable for the user's residence.

6. A user vehicle-related information acquisition unit acquires user-specific information, which is user-specific information related to the vehicle, from a user-specific information storage unit. A user query input unit that receives vehicle-related queries entered by the user via the user terminal operated by the user, A response information generation unit that, based on the query, generates a prompt which is an input sentence to a large-scale language model and transmits it to the large-scale language model, receives an original response to the prompt from the large-scale language model, and generates response information to be transmitted to the user terminal based on the original response, and, based on the user-specific information, decreases or increases, to a certain extent, at least one of the range of vocabulary, amount of information, and type of information that constitute the response information, corresponding to at least one of the amount of knowledge the user has about vehicles, the user's vehicle ownership history, the user's vehicle usage status, and the user's vehicle ownership status, A response information output unit that transmits the aforementioned response information to the user. A chatbot device characterized by having the following features.

Images