Information provision device, information provision method, program, and storage medium

Abstract

A server device 2 functions as an information provision device, and comprises an instruction information generation unit, an acquisition unit, and a notification unit. The instruction information generation unit generates instruction information including content inquiring about a recommended spot for an occupant of a vehicle of interest on the basis of at least one of a current situation relating to a movement state of the vehicle of interest or profile information of the occupant, and passenger information relating to a passenger accompanying a user. The acquisition unit acquires response information output from an LLM in response to the instruction information being input to the LLM. The notification unit notifies, via a vehicle terminal 1, the occupant of a recommended spot included in the response information and a reason for selecting the recommended spot for the passenger.

Description

Information Providing Device, Information Providing Method, Program, and Storage Medium

[0001] The present disclosure relates to the provision of information on recommended spots.

[0002] Conventionally, technologies for presenting information on spots recommended to users have been known. For example, Patent Document 1 discloses a search system that searches for spots existing within a search range dynamically changed along a route based on information on the current location and the route, and outputs information on the searched spots.

[0003] Japanese Unexamined Patent Application Publication No. 2021-12194

[0004] Focusing on the versatility of large language models, it is also conceivable to use a large language model to determine recommended spots. On the other hand, the answers of large language models may include information with low reliability. For example, when asking a large language model for recommended spots, there is a risk that non-existent spots may be erroneously output as recommended spots.

[0005] Also, as another independent problem, in a system that provides recommended spots to a user, the user does not understand why the spots recommended by the system were selected, and the user may feel resistance to immediately accepting the recommended spots provided by the system. Also, as another independent problem, when there is a passenger traveling with the user, it is necessary to consider selecting spots that the passenger can also accept.

[0006] One of the objectives of the present disclosure is to provide an information providing device capable of suitably providing information on spots in view of at least one of the above-described problems.

[0007] The invention described in claim 1 is an information providing device comprising: an instruction information generation unit that generates instruction information including a request for a recommended spot based on the current status of the moving state of a mobile body or at least one of the user's profile information and passenger information relating to the user and passengers riding in the mobile body; an acquisition unit that acquires response information output from a large-scale language model by inputting the instruction information into the large-scale language model; and a notification unit that notifies the occupants of the mobile body of the spot and the reason for selecting the spot for the passengers based on the response information.

[0008] The invention described in claim 11 is a control method for providing information performed by a computer, comprising: an instruction information generation step of generating instruction information that includes content asking for recommended spots based on the current status of the movement state of a mobile body or at least one of user profile information and passenger information relating to the user and passengers riding in the mobile body; an acquisition step of acquiring response information output from a large-scale language model by inputting the instruction information into the large-scale language model; and a notification step of notifying the occupants of the mobile body of the spots and the reasons for selecting those spots for the passengers based on the response information.

[0009] Furthermore, the invention described in claim 12 is a program that causes a computer to function as a notification unit that generates instruction information including content asking for recommended spots based on the current status of the moving state of a mobile body or at least one of the user's profile information and passenger information relating to the user and passengers riding in the mobile body; an acquisition unit that acquires response information output from a large-scale language model by inputting the instruction information into the large-scale language model; and a notification unit that notifies the occupants of the mobile body of the spots and the reasons for selecting those spots for the passengers based on the response information.

[0010] This shows an example of the configuration of the spot display system according to the embodiment. This shows an example of the general configuration of the vehicle terminal. (A) This shows an example of the general configuration of the server device. (B) This is an example of the general configuration of the model execution device. This is an example of a flowchart executed by the server device regarding the notification of recommended spots. This is an example of text representing instruction information. This is an example of text representing response information obtained based on the instruction information shown in Figure 5. This is an example of a user setting screen. This is the first example of text represented by the response information in Modification 1. This is the second example of text represented by the response information in Modification 1. (A) This is an example of text represented by the instruction information when the target vehicle is traveling on a road section where the user's travel history is below a predetermined number of times, or when it is traveling on a congested section. (B) This is an example of text represented by the instruction information when the user's passenger is a child. (C) This is an example of text represented by the instruction information when the user is traveling on a road section that is frequently used. This is the first example of the configuration of the spot display system in Modification 5. This is the second example of the configuration of the spot display system in Modification 5. This is an example of a flowchart executed by the server device in Modification 6. This is an example of a flowchart executed by the server device in Modification 7. This is an example of text representing instruction information generated by the server device in Modification 8. This is an example of text representing response information generated based on the instruction information shown in Figure 15. This is an example of text representing instruction information generated by the server device in Modification 9. This is an example of text representing question information generated based on the instruction information shown in Figure 17. This is an example of text representing response information generated based on answer information showing the answer to the question information shown in Figure 18. This is an example of a flowchart executed by the server device in Modification 9.

[0011] In one preferred embodiment of the present invention, the information providing device includes: an instruction information generation unit that generates instruction information including a request for recommended spots based on at least one of the current status regarding the movement state of a mobile body or user profile information and passenger information relating to the user and passengers riding in the mobile body; an acquisition unit that inputs the instruction information into a large-scale language model and acquires response information output from the large-scale language model; and a notification unit that, based on the response information, notifies the occupants of the mobile body of the spots and the reasons for selecting those spots for the passengers. In this embodiment, the information providing device can notify the user of recommended spots along with the reasons for selecting them for the passengers, thereby facilitating the smooth decision-making process between the user and passengers regarding which spots to visit without disputes.

[0012] In one embodiment of the information providing device described above, the instruction information generation unit generates instruction information, including the specification of the format of the response output by the large-scale language model, based on the attributes of the occupant or the movement history of the mobile body, and the notification unit notifies the occupant of the reason for the selection expressed in the response information according to the format. In this embodiment, the information providing device can notify the occupant of the reason for the selection in a format corresponding to the occupant's attributes or the movement history of the mobile body.

[0013] In another embodiment of the information providing device described above, the instruction information generation unit generates instruction information specifying that the reason for selection is concise when the occupants include children or when the vehicle is traveling on a road section whose travel history is less than or equal to a predetermined number of times, and the notification unit notifies the occupants of the reason for selection, which is concisely expressed in the response information. According to this embodiment, the information providing device can notify the occupants of the reason for selection, which is concisely expressed, when the occupants include children or when the vehicle is traveling on an unfamiliar road section.

[0014] In another embodiment of the information providing device described above, the instruction information generation unit generates instruction information that further includes content inquiring about the reputation of the spot, the acquisition unit acquires the response information that includes information about the reputation, and the notification unit notifies the crew of the information about the reputation as one of the reasons for the selection. In this embodiment, the information providing device can notify the crew of the reasons for the selection, including information about the reputation.

[0015] In another embodiment of the information providing device described above, the passenger information includes information indicating the relationship between the user and the passenger. In this embodiment, the information providing device can notify the occupant of the reasons for the selection, taking into account the relationship between the user and the passenger.

[0016] In another embodiment of the information providing device described above, the passenger information includes the passenger's profile information. In this embodiment, the information providing device can notify the occupant of the reasons for the selection, taking into account the passenger's profile information.

[0017] In another embodiment of the information providing device described above, the passenger information is generated based on user input when a passenger is detected when boarding the mobile vehicle. In this embodiment, the information providing device can suitably generate passenger information.

[0018] In another embodiment of the information providing device described above, the instruction information generation unit generates instruction information that further includes instructions for generating questions for selecting a spot based on at least one of the current situation or the profile information, and the acquisition unit inputs the instruction information into a large-scale language model to acquire questions output from the large-scale language model, and then supplies the crew's answers to the questions to the large-scale language model to acquire response information generated by the large-scale language model based on the answers. In this embodiment, the information providing device can confirm the user's current feelings through appropriate questions according to the current situation or profile, and select spots with higher accuracy.

[0019] In another embodiment of the information providing device described above, the instruction information generation unit generates instruction information that includes an instruction to generate a predetermined number of options, each assigned a number, as candidates for the answer. In this embodiment, the information providing device can generate questions that can be answered by number and select spots according to the answers.

[0020] In another embodiment of the information providing device described above, the instruction information generation unit generates instruction information that includes an instruction to generate a question to ascertain the current feelings of the crew member. In this embodiment, the information providing device can select a spot that takes into account the current feelings of the crew member.

[0021] In another preferred embodiment of the present invention, a computer-based information provision method comprises: an instruction information generation step of generating instruction information that includes a request for recommended spots based on the current status of the moving state of a mobile body or at least one of user profile information and passenger information relating to the user and passengers riding in the mobile body; an acquisition step of inputting the instruction information into a large-scale language model and acquiring response information output from the large-scale language model; and a notification step of notifying the occupants of the mobile body of the spots and the reasons for selecting those spots for the passengers based on the response information. By performing this information provision method, the computer can facilitate the user and passengers in smoothly deciding on spots to visit.

[0022] In yet another embodiment of the present invention, the program includes an instruction information generation unit that generates instruction information including a request for recommended spots based on the current status of the moving state of the mobile body or at least one of the user's profile information and passenger information relating to the user and passengers riding in the mobile body; an acquisition unit that inputs the instruction information into a large-scale language model and acquires response information output from the large-scale language model; and a notification unit that, based on the response information, notifies the occupants of the mobile body of the spots and the reasons for selecting those spots for the passengers. By executing this program, the computer can facilitate the user and passengers in smoothly deciding on spots to visit. Preferably, the program is stored in a storage medium.

[0023] Preferred embodiments of the present invention will be described below with reference to the drawings.

[0024] (1) System Configuration Diagram 1 shows an example of the configuration of the spot suggestion system according to the embodiment. The spot suggestion system is a system that provides information on recommended spots and comprises a vehicle terminal 1, a server device 2, and a model execution device 3. Here, "spot" refers to any place that can be set as a destination (including places to stop by, the same applies hereinafter), and may be any facility or a place such as a tourist spot.

[0025] Vehicle terminal 1 moves with the vehicle (also called the "target vehicle") in which the user of this system is riding, and functions as a user interface that receives input from the user and notifies the user of necessary information. Vehicle terminal 1 communicates data with server device 2 and provides driving assistance to the user, who is an occupant of the target vehicle. An example of driving assistance in this embodiment is the presentation of recommended spots (also called "recommended spots") to the user, and vehicle terminal 1 notifies the user of information regarding recommended spots by display and / or by voice. In this case, as will be described later, vehicle terminal 1 notifies the user of information such as recommended spots determined using a Large Language Model (LLM) and the reasons for recommending (selection reasons) for the recommended spots. In addition to the presentation of recommended spots, the above-mentioned driving assistance may also include route searching from the current location to the destination, route guidance, vehicle control related to automated driving to the destination, and other optional driving assistance.

[0026] Vehicle terminal 1 may be a navigation device installed in the target vehicle that provides route guidance to a set destination, or it may be a user's mobile terminal such as a smartphone on which an application that implements route guidance and other functions is installed. Vehicle terminal 1 may also be incorporated into the target vehicle. Vehicle terminal 1 is an example of an "information providing device". The target vehicle or user is an example of a "mobile entity".

[0027] Server device 2 generates information necessary for the vehicle terminal 1 to perform driving assistance in response to user requests, and supplies the generated information to the vehicle terminal 1, causing the vehicle terminal 1 to perform driving assistance to the user. For example, if server device 2 receives information from the vehicle terminal 1 requesting a recommended spot to present to the user, it supplies instruction information to the model execution device 3 instructing a response by the LLM. The instruction information functions as a prompt input to the LLM. Server device 2 then receives response information regarding the recommended spot from the model execution device 3 as a response to the instruction information and supplies the received response information to the vehicle terminal 1. The response information includes the recommended spot and the reason for selecting the recommended spot. In addition, if the vehicle terminal 1 makes a request for driving assistance other than a request for a recommended spot, server device 2 generates information corresponding to that request and transmits that information to the vehicle terminal 1. Examples of requests for driving assistance other than a request for a recommended spot include requests to perform route searching to a destination specified by the user, and requests for information regarding route guidance along a set route.

[0028] Server device 2 may be a system (cloud system) consisting of multiple devices or computers that collaborate using cloud computing technology or the like.

[0029] Model execution device 3 is a device that executes LLM and stores model information for constructing a machine-learned LLM. When model execution device 3 receives instruction information regarding recommendation spots from server device 2, it inputs the instruction information into the LLM configured based on the model information and sends the response information output by the LLM to server device 2. Examples of LLM include GPT (Generative Pre-trained Transformer), which outputs a sentence containing the input string by predicting the string with the highest probability following the input string, and ChatGPT, which is based on GPT.

[0030] (2) Device configuration diagram 2 shows an example of the schematic configuration of the vehicle terminal 1. The vehicle terminal 1 mainly includes a communication unit 11, a storage unit 12, an input unit 13, a control unit 14, a sensor group 15, a display unit 16, and a sound output unit 17. Each element within the vehicle terminal 1 is interconnected via a bus line 10.

[0031] The communication unit 11 communicates data with the server device 2 based on the control of the control unit 14. For example, based on the control of the control unit 14, the communication unit 11 transmits information regarding the movement status of the target vehicle (including the history of the target vehicle's location) identified based on the data output by the sensor group 15 to the server device 2, and receives information from the server device 2 necessary for controlling the output of the display unit 16 and the sound output unit 17.

[0032] The storage unit 12 is composed of various types of memory, including RAM (Random Access Memory), ROM (Read Only Memory), and non-volatile memory (including hard disk drives, flash memory, etc.). The storage unit 12 stores programs and software (including applications installed on the vehicle terminal 1) that enable the vehicle terminal 1 to perform predetermined processes. The aforementioned applications may be any applications that provide content (including driving assistance) to the user on the vehicle terminal 1. The storage unit 12 is also used as working memory for the control unit 14. The programs executed by the vehicle terminal 1 and other information may be stored in external devices other than the storage unit 12 that communicate with the vehicle terminal 1 (including server devices), a storage medium that can be attached to or detached from the vehicle terminal 1, or any other storage medium.

[0033] The input unit 13 is a user interface that accepts user input, and examples of the input unit 13 include buttons, touch panels, remote controllers, and voice input devices. The display unit 16 displays information based on the control of the control unit 14. Examples of the display unit 16 include displays and projectors. The sound output unit 17 outputs sound based on the control of the control unit 14. Examples of the sound output unit 17 include speakers.

[0034] The sensor group 15 includes various sensors that perform sensing of the state of the target vehicle or the environment outside the vehicle. The sensor group 15 has an external sensor 18 and an internal sensor 19. The external sensor 18 is one or more sensors for recognizing the surrounding environment of the target vehicle, such as an external camera, lidar, radar, ultrasonic sensor, infrared sensor, or sonar. The internal sensor 19 is a sensor for positioning the vehicle, such as a GNSS (Global Navigation Satellite System) receiver, gyro sensor, IMU (Internal Measurement Unit), vehicle speed sensor, or a combination thereof. Furthermore, the internal sensor 19 may also include sensors for detecting the conditions inside the target vehicle, such as a camera that takes pictures of the interior or seating sensors for each seat. Furthermore, the sensor group 15 only needs to include sensors that allow the control unit 14 to directly or indirectly derive the position of the target vehicle from the output of the sensor group 15 (i.e., by performing estimation processing).

[0035] The control unit 14 includes a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), and other components, and controls the entire vehicle terminal 1.

[0036] Furthermore, the processing performed by the control unit 14 is not limited to being implemented by software through a program, but may also be implemented by a combination of hardware, firmware, and software. Additionally, the processing performed by the control unit 14 may be implemented using a user-programmable integrated circuit, such as an FPGA (Field-Programmable Gate Array) or a microcontroller. In this case, the program executed by the control unit 14 in this embodiment may be implemented using this integrated circuit.

[0037] The configuration of the vehicle terminal 1 shown in Figure 2 is an example, and various modifications may be made to the configuration shown in Figure 2. For example, at least one of the input unit 13, the display unit 16, and the sound output unit 17 may be provided inside the target vehicle as an external device to the vehicle terminal 1, and the generated signals may be supplied to the vehicle terminal 1. Also, at least some of the sensors in the sensor group 15 may be sensors installed in the target vehicle. In this case, the vehicle terminal 1 may acquire information output by the sensors installed in the target vehicle from the target vehicle based on a communication protocol such as CAN (Controller Area Network).

[0038] Figure 3(A) shows an example of the schematic configuration of the server device 2. The server device 2 mainly consists of a communication unit 21, a storage unit 22, and a control unit 24. Each element within the server device 2 is interconnected via a bus line 20.

[0039] The communication unit 21 includes a communication antenna and a communication transceiver, and performs data communication with the vehicle terminal 1 and other devices such as the model execution device 3 based on the control of the control unit 24.

[0040] The storage unit 22 is composed of various types of memory, such as RAM, ROM, and non-volatile memory. The storage unit 22 stores programs for the server device 2 to perform predetermined processes. The storage unit 22 is also used as working memory for the control unit 24. The programs executed by the server device 2 may be stored in storage media other than the storage unit 22.

[0041] Furthermore, the storage unit 22 stores map information 5. Map information 5 records various data necessary for route guidance. Map information 5 is a map database necessary for displaying a map based on a predetermined location such as the current location. Map information 5 includes, for example, a road database (DB: DataBase) that represents the road network using a combination of nodes and links, and a spot DB, which is a database of spots that are candidates for destinations (including places to stop by). The spot DB has records that show attribute information for each spot. The attribute information for a spot includes location information, name, type (category), a chain store flag indicating whether or not it is a chain store, the chain store name if it is a chain store, business information, an audio file for explaining the spot via voice output, photographic data of the spot, and word-of-mouth information for the spot. The word-of-mouth information for a spot may be obtained by the server device 2 from multiple users who input word-of-mouth information for each spot, or it may be collected by the server device 2 by patrolling arbitrary websites under the control of the control unit 24. Business information includes information on business days or holidays, and may also include information indicating business hours.

[0042] Map information 5 is updated and regularly maintained under the control of the control unit 14, based on information received by the communication unit 11 from a management server (not shown). As a result, records of closed (unavailable) spots are removed from the spot database of map information 5, and records of existing spots containing the latest business information are registered in the spot database.

[0043] Furthermore, the storage unit 22 stores user-related information 6. User-related information is arbitrary information that the server device 2 can collect regarding the user of the target vehicle. Examples of user-related information include historical information such as the travel history of places the user has traveled using the target vehicle in the past and the visit history (number of visits) of places the user has visited in the past, setting information set by the user regarding recommended spots, and profile information regarding the user's profile. User-related information is linked, for example, to a user ID that identifies the user.

[0044] The control unit 24 includes a CPU, a GPU, etc., and controls the entire server device 2. For example, when the control unit 24 receives information requesting a recommended spot from the vehicle terminal 1 via the communication unit 21, it executes a process of providing the recommended spot to the vehicle terminal 1. In another example, when the control unit 24 receives a route setting request specifying a destination, a current position, and route search conditions from the vehicle terminal 1 via the communication unit 21, it performs a route search from the current position included in the setting request to the destination by referring to the road DB of the map information 5. In this case, the control unit 24 may perform the route search using an arbitrary route search algorithm such as Dijkstra's algorithm. Then, the control unit 24 transmits the search results of one or more routes to the vehicle terminal 1, and when information indicating the route specified by the user is supplied from the vehicle terminal 1, it sets the specified route as the route of the target vehicle and stores route information indicating the set route in the storage unit 22.

[0045] The control unit 24 functions as a computer or the like that executes a program. Note that the processes executed by the control unit 24 are not limited to being realized by software based on a program, and may be realized by any combination of hardware, firmware, and software. The control unit 24 functions as a "route acquisition unit", "extraction unit", "instruction information generation unit", "acquisition unit", "notification unit", and a computer or the like that executes a program.

[0046] Here, supplementary explanation will be given about the user-related information 6. The server device 2 may generate the user-related information 6 based on the input information of the user acquired via the vehicle terminal 1, or may generate the user-related information 6 based on the data output by the sensor group 15 of the vehicle terminal 1. For example, the server device 2 generates the above-described history information based on the position information supplied from the vehicle terminal 1. Examples of generating setting information and profile information based on the input information of the user acquired via the vehicle terminal 1 will be described later.

[0047] Note that the configuration of the server device 2 shown in FIG. 3(A) is an example, and various changes may be made to the configuration shown in FIG. 3(A).

[0048] Figure 3(B) shows an example of the schematic configuration of the model execution device 3. The model execution device 3 mainly includes a communication unit 31, a storage unit 32, and a control unit 34. Each element within the model execution device 3 is interconnected via a bus line 30.

[0049] The communication unit 31 includes a communication antenna and a communication transceiver, and performs data communication with other devices such as the vehicle terminal 1 and the server device 2 based on the control of the control unit 34.

[0050] The storage unit 32 is composed of various memories such as a RAM, a ROM, and a non-volatile memory. A program for the model execution device 3 to execute predetermined processing is stored in the storage unit 32. Also, the storage unit 32 is used as a working memory for the control unit 34.

[0051] Further, the storage unit 32 stores the model information 7 of the learned LLM executed by the model execution device 3. The model information 7 includes learned parameters necessary for constructing the LLM. The model information 7 includes, for example, various parameters of the machine-learned deep learning model such as the layer structure, the neuron structure of each layer, the number of filters and the filter size in each layer, and the weights of each element of each filter.

[0052] The control unit 34 includes a CPU, a GPU, etc., and controls the entire model execution device 3. The control unit 34 functions as a computer or the like that executes a program. Note that the processing executed by the control unit 34 is not limited to being realized by software according to a program, and may be realized by any combination of hardware, firmware, and software.

[0053] Note that the configuration of the model execution device 3 shown in Figure 3(B) is an example, and various modifications may be made to the configuration shown in Figure 3(B).

[0054] (3) Notification of Recommended Spots Next, the process for notifying users of recommended spots will be explained. In general terms, the server device 2 extracts candidate recommended spots from the map information 5, selects a recommended spot from the extracted candidates, and generates instruction information instructing it to provide a reason. The server device 2 then receives response information output by the LLM from the model execution device 3 based on the generated instruction information and supplies this response information to the vehicle terminal 1. In this way, the recommended spot, along with the reason for its selection, is appropriately notified to the user via the vehicle terminal 1.

[0055] Figure 4 shows an example of a flowchart that server device 2 executes regarding the notification of recommended spots.

[0056] First, the server device 2 receives information requesting a recommended spot from the vehicle terminal 1 (step S10). In this case, in the first example, when the vehicle terminal 1 detects user input requesting a recommended spot from the user, it sends information requesting a recommended spot to the server device 2. In the second example, when the vehicle terminal 1 detects user input requesting route setting, it sends information requesting a recommended spot to the server device 2 as a candidate destination. In the third example, the vehicle terminal 1 sends information requesting a recommended spot to the server device 2 each time the target vehicle moves a predetermined distance. In the fourth example, the vehicle terminal 1 sends information requesting a recommended spot to the server device 2 each time a predetermined time (e.g., 2 hours) has elapsed. In the fifth example, the vehicle terminal 1 sends information requesting a recommended spot to the server device 2 when the target vehicle reaches a predetermined area (e.g., within 5 km of the destination). In the sixth example, vehicle terminal 1 sends information requesting a recommended spot to server device 2 when the target vehicle is caught in a traffic jam that meets predetermined criteria.

[0057] Next, the server device 2 determines whether or not a route has been set for the target vehicle (step S11). In this case, the server device 2 determines whether or not a route for the target vehicle has already been set based on the request from the vehicle terminal 1. If it determines that a route for the target vehicle has been set (step S11; Yes), the server device 2 obtains route information indicating the location of the set route (step S12). For example, if the server device 2 sets a route for the target vehicle based on a route setting request from the vehicle terminal 1, it stores route information indicating the set route in the storage unit 22, and reads the route information from the storage unit 22 in step S12. On the other hand, if a route for the target vehicle has not been set (step S11; No), the server device 2 proceeds to step S13.

[0058] Next, the server device 2 extracts candidate recommended spots (also called "candidate spots") from the spot DB of the map information 5 (step S13). In this case, if route information was obtained in step S12, the server device 2 extracts spots from the spot DB that are around the current location of the target vehicle and along the route indicated by the route information, and designates the extracted spots as candidate spots. In this case, for example, the server device 2 extracts spots that are within a first distance from the location indicated by the latest location information of the target vehicle supplied from the vehicle terminal 1, and whose shortest distance to the route is within a second distance. The first and second distances mentioned above are, for example, predetermined distances. If no route has been set and route information was not obtained in step S12, the server device 2 extracts spots that are around the current location of the target vehicle from the spot DB, and designates the extracted spots as candidate spots.

[0059] Next, the server device 2 acquires current status information and user-related information 6 that indicate the current status of the movement of the target vehicle (step S14).

[0060] Here, examples of current information include driving information of the target vehicle, weather information for the area the target vehicle is traveling in, and passenger information. Driving information is information related to the driving of the target vehicle, such as the current location of the vehicle, the set destination, and the departure point. Passenger information is information about passengers, such as whether there are passengers and information about the attributes of the passengers. In addition, current information may include information indicating various situations occurring inside or around the target vehicle that can be detected by sensors, etc. Server device 2 generates driving information based on the location information of the target vehicle supplied from vehicle terminal 1, the departure point (current location during route search) and destination information specified by vehicle terminal 1 during route search, etc. Server device 2 may also obtain weather information from a server device that distributes weather information based on the current location of the target vehicle. Furthermore, server device 2 may receive images generated by a camera that photographs the interior of the target vehicle from vehicle terminal 1, analyze the images to identify the number of passengers and the attributes of each passenger (age, gender, etc.), and generate passenger information based on the identification results.

[0061] The server device 2 then generates instruction information asking for the recommendation spots and the reasons for selecting the recommendation spots, and transmits the generated instruction information to the model execution device 3 (step S15). The instruction information is text information asking for the recommendation spots and the reasons for selecting the recommendation spots, and is used as prompts for the LLM. In addition, the current information and user-related information 6 are reflected in the instruction information as information that the LLM should refer to when selecting recommendation spots. Specific examples of the instruction information will be described later.

[0062] Next, the server device 2 receives the response information output by the LLM based on the instruction information from the model execution device 3 and transmits the received response information to the vehicle terminal 1 (step S16). In this case, the model execution device 3 inputs the instruction information received from the server device 2 as a prompt to the LLM and transmits the response information output by the LLM in response to the input of the instruction information to the server device 2. The response information is text information that indicates the answer to the question indicated by the instruction information and includes recommended spots and the reasons for selecting the recommended spots. The vehicle terminal 1, having received the response information from the server device 2, then notifies the user of the recommended spots and the reasons for selecting the recommended spots included in the response information.

[0063] Then, when the server device 2 receives information from the vehicle terminal 1 indicating user input specifying one of the recommended spots included in the response information, it sets the specified recommended spot as the destination (including stops) of the target vehicle and performs a route search to that recommended spot (step S17). After that, the server device 2 executes processing related to driving assistance (such as route guidance) for the target vehicle to travel along the searched route.

[0064] Here, we will provide a supplementary explanation regarding step S13. In step S13, the server device 2 extracts candidate spots from the map information 5 as candidate spots for LLM to select as recommended spots. Here, the map information 5 includes a highly reliable spot DB that is regularly maintained, and the server device 2 extracts existing spots from such a highly reliable spot DB as candidate spots. Generally, when a prompt is input to the LLM asking for spots that meet predetermined conditions, the spots output by the LLM may include spots that no longer exist. Therefore, the server device 2 extracts existing, highly reliable candidate spots from the spot DB of the map information 5 and has the LLM select a recommended spot from these candidate spots.

[0065] In a preferred example in step S13, the server device 2 may filter candidate spots to extract only those spots that are currently in operation, based on the business information contained in the spot DB. For example, the server device 2 refers to the business information contained in the spot DB and identifies the days when each spot extracted from the spot DB based on the current location, etc., in step S13 is closed. Then, if the server device 2 determines that the day on which the recommended spot is selected (i.e., the current date) falls on a day when the spot is closed, it excludes the spot that falls on a closed day from the candidate spots. In other words, the server device 2 extracts only spots that are not closed at the time the recommended spot is selected as candidate spots. This ensures that the server device 2 does not select a spot as a recommended spot that is unavailable to the user even if they visit it.

[0066] In a more preferred example, the server device 2 may extract only spots that the target vehicle can reach within business hours as candidate spots, and exclude spots that are not expected to be reachable by the target vehicle within business hours. In this case, the server device 2, for example, refers to the business information contained in the spot DB and identifies the business hours of each spot extracted from the spot DB based on the current location, etc., in step S13. Then, if the server device 2 determines that the estimated arrival time to a spot does not fall within the business hours, it excludes that spot from the candidate spots. That is, the server device 2 extracts spots as candidate spots where the estimated arrival time does not fall on a closed day or outside of business hours. This ensures that the server device 2 does not select spots that users cannot use even if they visit them as recommended spots. The server device 2 may calculate the estimated arrival time using any method. For example, the server device 2 may calculate the estimated arrival time to each spot by assuming that the target vehicle moves at a predetermined speed, or it may use the time after a predetermined time has elapsed from the current time instead of the estimated arrival time.

[0067] Figure 5 shows an example of text representing instruction information generated by the server device 2. The text shown in Figure 5 includes a question block 51, a candidate spot block 52, a current information block 53, and a user profile block 54.

[0068] Question block 51 indicates that the current status of the target vehicle's user, the user's profile, and the number of past visits to each candidate spot should be taken into consideration, that a predetermined number (two in this case) of recommended spots should be selected from the candidate spots, and that the reasons for the selection of the recommended spots should be attached. The server device 2 may generate question block 51 using, for example, standard text information stored in the storage unit 22.

[0069] In candidate spot block 52, six candidate spots are shown along with the number of past visits for each user. The server device 2 identifies the number of past visits for each candidate spot extracted from map information 5 by referring to the history information of user-related information, and lists the pairs of candidate spots and the number of visits for each user in candidate spot block 52.

[0070] The current status information block 53 shows the current status of the vehicle's movement. Specifically, the current status information block 53 shows the current location, departure point, destination, current weather, and passenger attributes (in this case, information on adults / infants). Based on the current status information, which includes driving information, weather information, and passenger information, the server device 2 generates the current status information block 53 showing the current location, departure point, destination, current weather, and passenger attributes. In this case, the desired arrival time at the destination is specified in advance by user input, and the server device 2 also includes the desired arrival time based on user input in the current status information block 53.

[0071] The user profile block 54 displays the user's profile. The server device 2 generates the user profile block 54 based on the profile information included in the user-related information.

[0072] Figure 6 is an example of text representing the response information that the server device 2 obtains from the model execution device 3 based on the instruction information shown in Figure 5. The server device 2 transmits instruction information corresponding to the text shown in Figure 5 to the model execution device 3, and the model execution device 3 inputs the instruction information received from the server device 2 into the LLM, thereby transmitting the response information output by the LLM back to the server device 2.

[0073] The response information shown in Figure 6 represents the answer to the question indicated by the instruction information in Figure 5. Based on the instruction information, it includes the recommended spots "Kawagoe B-ya" and "Kawagoe Bakery D" from the candidate spots, along with the reasons for their selection. Here, LLM selects two recommended spots by referring to the information shown in the candidate spot block 52, current information block 53, and user profile block 54, based on the specifications in the question block 51 of Figure 5. Specifically, LLM selects two spots from the candidate spots listed in the candidate spot block 52 that have the lowest number of visits and that match the preferences based on the user's profile. In addition, LLM generates text information representing the reasons for selecting each selected spot based on the specifications in the question block 51 of the instruction information, and includes this text information in the response information.

[0074] Vehicle terminal 1 receives response information shown in Figure 6 from server device 2, and based on the received response information, displays and / or outputs the recommended spot and the reason for its selection via voice. As a result, vehicle terminal 1 can notify the user of carefully selected recommended spots, along with the reason for their selection, from candidate spots extracted from highly reliable map information 5, taking into account the user's profile, current situation, visit history, etc.

[0075] Here, we will provide a supplementary explanation regarding the effectiveness of notifying the user of the selection reasons. Even if a recommended spot is selected considering the user's characteristics, the user may feel resentment towards something recommended by the system if they do not understand the sufficient reasons for the recommendation. Taking this into consideration, server device 2 generates instruction information that asks for the selection reasons in addition to the recommended spot. This allows the selection reasons for the recommended spot to be obtained from LLM and notified to the user, thereby effectively suppressing user resentment.

[0076] Figure 7 shows an example of a user settings screen. Before requesting a recommendation spot from the server device 2, the vehicle terminal 1 displays the user settings screen shown in Figure 7 and accepts user input to generate user-related information used in selecting a recommendation spot. In the user settings screen of Figure 7, the vehicle terminal 1 has a target genre specification area 61 and a user profile specification area 62.

[0077] In the designated genre area 61, the vehicle terminal 1 displays a user interface for specifying the genre of the spot to be selected as a recommended spot. As an example, the vehicle terminal 1 provides checkboxes for each anticipated genre, allowing the user to select the genre for the recommended spot. While only the genre of dining establishments is shown here, genres of other types of facilities (e.g., entertainment facilities) may also be displayed and selectable in a similar manner.

[0078] In the user profile designation area 62, the vehicle terminal 1 displays a user interface that accepts input regarding the user's profile. Here, as an example, the vehicle terminal 1 accepts the user's preference level (in this case, three levels: love, like, dislike) for various fields (sweets, spicy food, exercise, history, etc.) that are useful in determining recommended spots.

[0079] Furthermore, on the user settings screen, the vehicle terminal 1 may accept input not only of the target genre of recommended spots and user profile, but also of any other information that can be used to select recommended spots (for example, passenger information).

[0080] Then, when the vehicle terminal 1 detects user input indicating completion of input (for example, the selection of a completion button, not shown), it sends input information indicating the content entered on the user settings screen to the server device 2. Based on the input information received from the vehicle terminal 1, the server device 2 generates user-related information representing the information specified by the user.

[0081] As described above, the server device 2 functions as an information providing device and includes an instruction information generation unit, an acquisition unit, and a notification unit. The instruction information generation unit generates instruction information that includes a request for recommended spots to the occupant, based on at least one of the current status regarding the movement status of the target vehicle or the profile information of the occupant of the target vehicle. The acquisition unit inputs the instruction information into the LLM and acquires the response information output from the LLM. The notification unit notifies the occupant of the recommended spots included in the response information and the reasons for selecting those recommended spots via the vehicle terminal 1. In this way, the server device 2 can notify the user of the reasons for selecting the recommended spots along with the recommended spots, and give the user a sense of satisfaction with the notified recommended spots.

[0082] Furthermore, the server device 2 functions as an information providing device and includes a route acquisition unit, an extraction unit, an instruction information generation unit, and a notification unit. The route acquisition unit acquires route information indicating the route of the target vehicle. The extraction unit extracts candidate spots that the target vehicle will visit from the map information 5 based on the route information. The instruction information generation unit generates instruction information that includes content instructing the occupant to select a recommended spot from the candidates, based on the current status regarding the movement status of the target vehicle or the profile information of the occupant of the target vehicle. The notification unit inputs the instruction information into the LLM and, based on the response information output from the LLM, notifies the occupant of the recommended spot selected from the candidates via the vehicle terminal 1. In this way, the server device 2 extracts candidate spots for recommended spots from the highly reliable map information 5 and specifies that a recommended spot be selected from these candidates. As a result, the server device 2 can notify the user of a recommended spot with guaranteed reliability.

[0083] (4) Modifications Next, modifications suitable for the above-described embodiments will be explained. The following modifications may be applied in combination to the above-described embodiments.

[0084] (Modification 1) Server device 2 may generate instruction information so that the response information includes user reviews indicating the evaluation of the recommended spot as the reason for selecting the recommended spot. For example, server device 2 generates instruction information that instructs to include user reviews of the selected recommended spot (which may be limited to positive reviews) as the reason for selection. User reviews are an example of information related to reputation.

[0085] Figure 8 shows a first example of the text represented by the response information acquired by the server device 2 in Modification 1. Here, the server device 2 generates instruction information that includes text information such as, "In the selection reason, please also introduce positive online reviews of the selected spot," and obtains the response information generated by the LLM based on the instruction information by having the LLM summarize the reviews about the selected recommended spot from the review information stored in the spot DB. At this time, the server device 2 may include multiple review information for each candidate spot in the instruction information and send it to the model execution device 3, or it may send the storage address of the review information for the candidate spots to the model execution device 3 and have the model execution device 3 refer to the review information. Alternatively, the server device 2 may periodically send the review information stored in the spot DB to the model execution device 3 and have the review information stored in the model execution device 3's storage unit 32 store the review information, and have the LLM refer to the stored review information and summarize it.

[0086] Figure 8 lists the reasons for selecting "Kawagoe B-ya," one of the selected recommended spots, with "word-of-mouth evaluation" listed as the third reason. Here, text information summarizing the word-of-mouth reviews of the recommended spot is generated as "word-of-mouth evaluation." In this way, by generating instruction information that instructs the server device 2 to introduce positive word-of-mouth reviews, it can obtain response information that includes word-of-mouth evaluation as the reason for selecting the recommended spot, and notify the user of a more convincing reason for selection.

[0087] Furthermore, server device 2 may generate instruction information that instructs the introduction of recent reviews. Figure 9 shows a second example of the text represented by the response information acquired by server device 2 in Modification 1. Here, server device 2 generates instruction information that includes text information such as, "In the selection reason, please introduce positive reviews written on the internet after January 1, 2024, including the date of posting," and acquires the response information generated by LLM based on this instruction information. In this case, for example, the selection reason for "Kawagoe B-ya," one of the recommended spots, includes the evaluation of a review that reads, "Review: A review from July 2024 gave a high rating, saying, 'I ordered the Kiku eel rice bowl. The plump eel in the Kanto style, soaked in sauce, was superb. The soup was also delicious. It was a taste worthy of being listed as one of the top 100 restaurants.'" In this way, server device 2 can include information on reputation based on reviews with posting dates newer than a predetermined date in the selection reason, and notify users of the reputation of recommended spots based on fresh reviews.

[0088] (Modification 2) The server device 2 may generate instruction information to change the content of the selection reason (i.e., the information introducing the recommended spot) based on the user's visit history to the recommended spot.

[0089] For example, server device 2 may generate instruction information that instructs the system to introduce reviews and other reputation information for recommended spots whose visit count is less than or equal to a predetermined number (e.g., 0 times). In this case, server device 2 may generate instruction information that includes text information such as, "Please introduce reviews of a recommended spot only if it is the first visit," and obtain response information generated by LLM based on this instruction information. In another example, server device 2 may generate instruction information that includes text information such as, "Please introduce the general evaluation of a recommended spot only if it is the first visit," and obtain response information generated by LLM based on this instruction information. This response information includes reviews introducing the recommended spot. In this case, the reviews are information introducing the recommended spot and are an example of introduction information.

[0090] In this way, the server device 2 can reduce user anxiety about a spot they are visiting for the first time, while suppressing excessive information provision about spots the user already knows.

[0091] (Modification 3) After receiving response information from the model execution device 3, the server device 2 may add information indicating the last visit date of the user to each recommended spot that is not the first visit, and notify the user of the recommended spot information including the visit date. In this case, the server device 2 refers to the history information of the user-related information 6 corresponding to the recommended spot included in the response information received from the model execution device 3, and identifies the last visit date of the user to each recommended spot.

[0092] In this way, the server device 2 according to Modification 3 can use historical information and other data not managed by LLM to add information about visit history and other data to information about recommended spots, thereby enabling it to notify users of more comprehensive information about recommended spots.

[0093] (Modification 4) The server device 2 may generate instruction information, including a specification of the format of the response to be output by the LLM, based on at least one of the movement history or the attributes of the crew.

[0094] Figure 10(A) is an example of the text represented by the instruction information generated by the server device 2 when the target vehicle is traveling on a road section where the user's travel history is below a predetermined number of times, or when it is traveling on a congested section.

[0095] In this case, the server device 2 generates instruction information that includes text information instructing the LLM to respond in the format of "Please explain in a short sentence." For example, in this case, the server device 2 identifies the travel history of a driving section identified based on the current location supplied from the target vehicle and the map information 5, by referring to user-related information 6, and generates instruction information requesting a concise reason for selection as described above if the travel history in the identified driving section is less than or equal to a predetermined number of times. In another example, the server device 2 determines that the driving section is a congested section based on the travel history of the driving section identified based on the current location supplied from the target vehicle and the map information 5, and traffic congestion information received from a server device that manages traffic congestion information, etc., and generates instruction information requesting a concise reason for selection as described above. The server device 2 then supplies the generated instruction information to the model execution device 3, causing the LLM to generate response information expressing the reason for selection in a short sentence, and the server device 2 notifies the user of the recommended spot along with a concise reason for selection via the vehicle terminal 1 based on this response information.

[0096] As shown in the example in Figure 10(A), the server device 2 can concisely inform the user of the reason for its selection in situations where the user needs to pay attention while driving, such as on unfamiliar road sections or in congested areas, thereby allowing the user to concentrate on driving.

[0097] Figure 10(B) shows an example of the text represented by the instruction information generated by the server device 2 when the user's passenger is a child.

[0098] In this example, because the server device 2 has a child among the passengers, it adds the text information "Please explain in simple terms that even a child can understand" to the instruction information. In this case, the server device 2 determines whether or not there is a child among the passengers based on the passenger information, and if it determines that there is a child, it generates the instruction information mentioned above. The server device 2 then supplies the generated instruction information to the model execution device 3, so that the LLM generates response information that expresses the selection reason in simple terms that even a child can understand, and the server device 2 notifies the user of the recommended spot and the selection reason via the vehicle terminal 1 based on this response information.

[0099] According to the example in Figure 10(B), if a child is among the passengers of the user, the server device 2 can notify the user of the reason for selecting the recommended spot in a way that all passengers, including the child, can understand.

[0100] Figure 10(C) shows an example of the text represented by the instruction information generated by the server device 2 when the user is driving on a road section that is frequently used by the user.

[0101] In this example, the server device 2 determines that the user is traveling on a road section that the user frequently travels on and adds the instruction "Prioritize selecting spots that have not been visited in the user's history" to the instruction information. In this case, the server device 2 identifies the travel history of the travel section identified based on the current location supplied from the target vehicle and the map information 5, by referring to the user-related information 6. If the travel history in the identified travel section exceeds a predetermined number of times, the server device 2 determines that the user is traveling on a road section that the user frequently travels on and generates the instruction information described above. The server device 2 then supplies the generated instruction information to the model execution device 3, causing the LLM to generate response information that designates spots with no visit history as recommended spots. Based on this response information, the server device 2 notifies the user of the recommended spots with no visit history via the vehicle terminal 1.

[0102] According to the example in Figure 10(C), when the user is driving on a road section they are familiar with, the server device 2 can prioritize recommending spots that the user has not yet visited.

[0103] (Modification 5) The spot display system is not limited to a configuration having a vehicle terminal 1, a server device 2, and a model execution device 3 as shown in Figure 1.

[0104] Figure 11 shows a first configuration example of the spot display system in this modified example. The spot display system shown in Figure 11 includes a vehicle terminal 1 and a server device 2A. The server device 2A functions as the server device 2 and model execution device 3 shown in Figure 1, and stores model information 7. The server device 2A then executes the LLM by referring to the model information 7 based on the generated instruction information and obtains the response information output by the LLM.

[0105] Figure 12 shows a second configuration example of the spot suggestion system in this modified example. The spot suggestion system shown in Figure 12 includes a vehicle terminal 1A and a model execution device 3. The vehicle terminal 1A functions as the vehicle terminal 1 and server device 2 shown in Figure 1, and stores map information 5 and user-related information 6, etc. When notifying the user of a recommended spot, the vehicle terminal 1A generates instruction information based on the current information and user-related information 6, etc., and supplies the generated instruction information to the model execution device 3. The vehicle terminal 1A then receives response information output by the LLM when the model execution device 3 inputs the instruction information into the LLM, and notifies the user of the recommended spot and the reason for its selection as indicated in the received response information.

[0106] Even with this modified configuration, the spot suggestion system can effectively notify users of information about recommended spots.

[0107] (Modification 6) The server device 2 may generate instruction information without extracting candidate spots.

[0108] Figure 13 shows an example of a flowchart executed by the server device 2 in modified example 6.

[0109] First, the server device 2 receives information requesting a recommended spot from the vehicle terminal 1 (step S20). Next, the server device 2 acquires at least one of the current status information and user-related information 6, which indicates the current status of the movement of the target vehicle (step S21). Then, based on at least one of the current status information and user-related information 6, the server device 2 generates instruction information asking for the recommended spot and the reason for selecting the recommended spot, and transmits the generated instruction information to the model execution device 3 (step S22). The server device 2 receives response information output by the LLM from the model execution device 3 based on the instruction information (step S23).

[0110] The server device 2 then notifies the user of the recommended spots and selection reasons included in the response information via the vehicle terminal 1 (step S24). In this case, the server device 2 transmits the response information acquired in step S23 to the vehicle terminal 1, causing the recommended spots and selection reasons included in the response information to be displayed and / or voiced to the vehicle terminal 1.

[0111] In this modified example as well, the server device 2 can suitably notify the user of information about recommended spots while reducing the user's resentment towards unilateral recommendations.

[0112] (Modification 7) The server device 2 may generate instruction information that does not require a reason for selecting the recommendation spot.

[0113] Figure 14 shows an example of a flowchart executed by the server device 2 in the modified example 7. Here, a destination and a route to the destination are set, and the server device 2 is assumed to have information regarding the set destination and route.

[0114] First, the server device 2 receives information requesting a recommended spot from the vehicle terminal 1 (step S30). Next, the server device 2 acquires route information (step S31). Next, the server device 2 extracts candidate spots from the spot DB of the map information 5 (step S32). Then, the server device 2 acquires at least one of the current status information and user-related information 6 that indicates the current status of the movement of the target vehicle (step S33). Then, based on at least one of the current status information and user-related information 6, the server device 2 generates instruction information instructing the selection of a recommended spot from the candidate spots and transmits the generated instruction information to the model execution device 3 (step S34). The server device 2 receives response information output by the LLM from the model execution device 3 based on the instruction information (step S35).

[0115] Then, the server device 2 notifies the user of the recommended spots included in the response information via the vehicle terminal 1 (step S36). In this case, the server device 2 transmits the response information acquired in step S23 to the vehicle terminal 1, causing the vehicle terminal 1 to output the recommended spots included in the response information by display and / or by sound.

[0116] In this modified example, the server device 2 can also cause the LLM to select recommended spots from candidate spots extracted from a highly reliable spot DB, and can suitably notify the user of information regarding highly reliable recommended spots.

[0117] (Modification 8) The server device 2 may generate instruction information that instructs the LLM to output selection reasons that allow it to recommend recommended spots to passengers other than the user if there are passengers in the target vehicle.

[0118] Figure 15 shows an example of text representing instruction information generated by the server device 2 in this modified example. The text shown in Figure 15 includes a question block 51, an additional instruction block 51A, a candidate spot block 52, a current information block 53, and a user profile block 54.

[0119] Question block 51 indicates that the current status of the target vehicle's user, the user's profile, and the number of past visits to each candidate spot should be taken into consideration, that a predetermined number (two in this case) of recommended spots should be selected from the candidate spots, and that the reasons for the selection of the recommended spots should be attached. The server device 2 may generate question block 51 using, for example, standard text information stored in the storage unit 22.

[0120] The additional instruction block 51A indicates that if there is a passenger accompanying the user in the target vehicle, the system should explain to the passenger, on behalf of the user, the reasons for selecting the recommended spot so that the passenger can also be recommended. For example, the server device 2 may include the additional instruction block 51A in the instruction information if it determines that there is a passenger based on the passenger information included in the current status information of the target vehicle.

[0121] In candidate spot block 52, seven candidate spots are shown along with the number of past visits for each user. The server device 2 identifies the number of past visits for each candidate spot extracted from map information 5 by referring to the history information of user-related information, and lists the pairs of candidate spots and the number of visits for each user in candidate spot block 52.

[0122] The current status information block 53 shows the current status of the vehicle's movement. Specifically, the current status information block 53 shows the current location, departure point, destination, current weather, and passenger attributes. The server device 2 generates the current status information block 53, which shows the current location, departure point, destination, current weather, and passenger attributes, based on the current status information, which includes driving information, weather information, and passenger information.

[0123] In the example shown in Figure 15, the relationship between each passenger and the user (here, "the user's mother" and "the user's father," respectively) is shown in the current information block 53 as an attribute of the passengers. Here, the vehicle terminal 1 may receive user input of passenger information when the user boards the target vehicle in order to obtain detailed attribute information of the passengers, such as information indicating the relationship between the passengers and the user, and supply the passenger information based on the received user input to the server device 2. For example, the server device 2 may receive images generated by a camera that takes pictures inside the target vehicle from the vehicle terminal 1, analyze the images to identify the number of passengers, and cause the vehicle terminal 1 to execute an output prompting the user to specify the relationship of each passenger to the user. In this case, the vehicle terminal 1 may make an audio output asking about the relationship of each passenger to the user and accept input (including voice input and touch input on the screen) specifying the relationship. In this case, the vehicle terminal 1 may store candidate relationships between the user and passengers, present the candidates to the user by voice or display, and allow the user to select the appropriate candidate. Examples of possible relationships include "father," "mother," "family" (which can be used instead of "father" or "mother"), "friend," "partner," and "business relationship" (which can be further subdivided into "boss," "subordinate," "colleague," etc.).

[0124] The user profile block 54 displays the user's profile. The server device 2 generates the user profile block 54 based on the profile information included in the user-related information.

[0125] In addition to the user's profile, the user profile block 54 may also show the profiles of passengers. In the example in Figure 15, the user profile block 54 may also include information showing the names and preferences of two people, the user's mother and the user's father. In this case, the server device 2 refers to the passenger information and includes the passenger's profile in the user profile block 54. As passenger information, family information showing attributes of family members (including father and mother) that the user has registered in advance may be used, or if the passenger is a user of the system, profile information (user-related information) associated with the passenger may be used. In the latter example, a user ID that the system uses to identify the passenger may be entered by the user when boarding the vehicle.

[0126] Figure 16 is an example of text representing the response information that the server device 2 obtains from the model execution device 3 based on the instruction information shown in Figure 15. The server device 2 transmits instruction information corresponding to the text shown in Figure 15 to the model execution device 3, and the model execution device 3 inputs the instruction information received from the server device 2 into the LLM, thereby transmitting the response information output by the LLM to the server device 2.

[0127] The response information shown in Figure 16 represents the answer to the question indicated by the instruction information in Figure 15, and includes the recommended spots "Kawagoe Bakery D" and "Eel C" selected from the candidate spots based on the instruction information, as well as the reason for selecting each recommended spot. Here, LLM selects the two recommended spots by referring to the information shown in the candidate spot block 52, the current information block 53, and the user profile block 54, based on the specifications in the question block 51 and additional instruction block 51A in Figure 15.

[0128] Here, LLM selects recommended spots based on the information shown in the additional instruction block 51A, considering whether or not it is a spot that can be recommended to passengers. In addition, based on the specifications in the instruction information question block 51 and the additional instruction block 51A, LLM generates text information that expresses the reason for selecting each selected spot and includes this text information in the response information. In this case, based on the specifications in the additional instruction block 51A, LLM generates text information that includes an explanation for passengers and includes this text information in the response information. Specifically, the "Reason for recommendation (explanation for passengers)" for the recommended spot "Kawagoe Bakery D" includes the explanation for passengers: "It's a great place for your mother and father to have a light cup of tea and take a break, so it's perfect for a break during travel in the heat." Also, the "Reason for recommendation (explanation for passengers)" for the recommended spot "Eel C" includes the explanation for passengers: "I've been there once before, so I feel at ease, and it's a calm restaurant that your parents will also enjoy."

[0129] Vehicle terminal 1 receives response information shown in Figure 16 from server device 2, and based on the received response information, displays and / or outputs the recommended spot and the reason for its selection via voice. As a result, vehicle terminal 1 can notify the user of carefully selected recommended spots, which are chosen from candidate spots extracted from highly reliable map information 5, taking into account the user's profile, current situation, visit history, etc., along with the reason for selection, including an explanation for passengers.

[0130] (Modification 9) The server device 2 may generate a prompt specifying that it should generate a question to confirm the user's current mood, and cause the LLM to output recommendation spots corresponding to the answers to the questions.

[0131] Figure 17 shows an example of text representing instruction information generated by the server device 2 in this modified example. The text shown in Figure 17 includes a question block 51, an additional instruction block 51B, a candidate spot block 52, a current information block 53, and a user profile block 54.

[0132] Question block 51 indicates that the current status of the vehicle's user, the user's profile, and the number of past visits to each candidate spot should be taken into consideration, that a predetermined number of recommended spots (two in this case) should be selected from the candidate spots, and that the reasons for the selection of the chosen recommended spots should be attached.

[0133] The additional instruction block 51B indicates that the server should ask the user appropriate questions to ascertain their current feelings, taking into account the current situation, the user's profile, and the number of past visits; that it should present the user with answer choices in response to the questions; and that the final recommended spot should be determined based on the user's answers to the questions. The server device 2 may generate the question block 51 and the additional instruction block 51B, for example, using standard text information stored in the storage unit 22.

[0134] In candidate spot block 52, seven candidate spots are shown along with the number of past visits for each user. The server device 2 identifies the number of past visits for each candidate spot extracted from map information 5 by referring to the history information of user-related information, and lists the pairs of candidate spots and the number of visits for each user in candidate spot block 52.

[0135] The current status information block 53 shows the current status of the vehicle's movement. Specifically, the current status information block 53 shows the current location, departure point, destination, current weather, and passenger attributes. The server device 2 generates the current status information block 53, which shows the current location, departure point, destination, current weather, and passenger attributes, based on the current status information, which includes driving information, weather information, and passenger information. Here, as in the example in Figure 15, the relationship between each passenger and the user is shown in the current status information block 53 as a passenger attribute.

[0136] The user profile block 54 displays the user's profile. The server device 2 generates the user profile block 54 based on the profile information included in the user-related information.

[0137] Figure 18 is an example of text representing the question information that the server device 2 obtains from the model execution device 3 based on the instruction information shown in Figure 17. The server device 2 transmits instruction information corresponding to the text shown in Figure 17 to the model execution device 3, and the model execution device 3 inputs the instruction information received from the server device 2 into the LLM, thereby transmitting the question information output by the LLM to the server device 2.

[0138] The question information shown in Figure 18 represents a question with multiple-choice options generated by LLM based on the instructions in additional instruction block 51B, taking into account the current situation, user profile, and past visit count. Here, the question is presented as, "Please choose the option that best describes your current mood," and three numbered options (i.e., answer candidates) are shown as possible answers. The first option is, "I want to take a break and eat something sweet in a cool place," the second option is, "I want to take a short walk while enjoying the history and scenery of the town," and the third option is, "I'm getting hungry, so I want to have a proper meal."

[0139] Server device 2 transmits the question information shown in Figure 18 to vehicle terminal 1, and outputs the question information to vehicle terminal 1 by display and / or voice. Vehicle terminal 1 generates answer information indicating the number of the option selected by the user for the outputted question information, based on the user's input, and transmits the answer information to server device 2. The user's input means for generating the answer information is arbitrary and may be voice input, touch input, or button input.

[0140] In this example, the user selects the third option as their answer, and the server device 2 receives the answer information specifying the third option as the answer from the vehicle terminal 1. The server device 2 then transmits the answer information received from the vehicle terminal 1 to the model execution device 3.

[0141] Figure 19 is an example of text representing response information that the server device 2 obtains from the model execution device 3 based on the response information received from the vehicle terminal 1. The server device 2 sends response information indicating the answer to the question generated by the LLM to the model execution device 3, and the model execution device 3 inputs the response information received from the server device 2 to the LLM, thereby sending the response information output by the LLM to the server device 2.

[0142] The response information shown in Figure 19 represents the LLM's response based on the instructions in Question Block 51 and the third option selected by the user as an answer to the question generated by the LLM: "I'm getting hungry, so I want to have a proper meal." Here, the LLM includes the recommended spots "Eel C" and "Udon Kawagoe Store" selected from the candidate spots, along with the reasons for selecting each recommended spot. Based on the specifications in Question Block 51, the LLM selects the two recommended spots by referring to the information shown in Candidate Spot Block 52, Current Information Block 53, and User Profile Block 54.

[0143] Here, LLM prioritizes selecting recommended spots where a substantial meal can be enjoyed, according to the third option selected by the user. Furthermore, based on the specifications in question block 51 of the instruction information, LLM generates text information representing the reason for selecting each spot and includes this text information in the response information. One of the selection reasons in this case is that the spot offers a substantial meal.

[0144] Vehicle terminal 1 receives response information shown in Figure 19 from server device 2, and based on the received response information, displays and / or outputs the recommended spot and the reason for its selection via voice. In this way, vehicle terminal 1 can notify the user of carefully selected recommended spots that accurately take into account the user's current mood, along with the reason for their selection.

[0145] Figure 20 shows an example of a flowchart executed by the server device 2 in this modified example.

[0146] First, the server device 2 receives information requesting recommended spots from the vehicle terminal 1 (step S40). Next, the server device 2 acquires at least one of the current status information and user-related information 6, which indicate the current status of the movement of the target vehicle (step S41). Then, based on at least one of the current status information and user-related information 6, the server device 2 generates instruction information that at least instructs the generation of questions for selecting recommended spots, and transmits the generated instruction information to the model execution device 3 (step S42). The server device 2 may also execute steps S12 (processing to acquire route information) and S13 (processing to extract candidate spots from map information 5) of the flowchart shown in Figure 4, and include candidate spots in the instruction information.

[0147] The server device 2 receives question information output by the LLM from the model execution device 3 based on the instruction information (step S43). Then, the server device 2 transmits the question information to the vehicle terminal 1 and receives answer information from the vehicle terminal 1 indicating the answer received by the vehicle terminal 1 based on the question information (step S44).

[0148] The server device 2 then transmits the response information to the model execution device 3 and receives the response information output by the LLM based on the instruction information and the response information from the model execution device 3 (step S45). The server device 2 then notifies the user of the recommended spots, etc., included in the response information via the vehicle terminal 1 (step S46). In this case, the server device 2 transmits the response information acquired in step S45 to the vehicle terminal 1, causing the vehicle terminal 1 to output information regarding the recommended spots included in the response information (which may include the reason for selection) by display and / or by voice.

[0149] In each of the embodiments described above, the program can be stored using various types of non-transitory computer-readable media and supplied to a control unit, which is a computer. Non-transitory computer-readable media include various types of tangible storage media. Examples of non-temporary computer-readable media include magnetic storage media (e.g., flexible disks, magnetic tapes, hard disk drives), magneto-optical storage media (e.g., magneto-optical disks), CD-ROMs (Read Only Memory), CD-Rs, CD-R / Ws, and semiconductor memory (e.g., mask ROMs, PROMs (Programmable ROMs), EPROMs (Erasable PROMs), flash ROMs, and RAMs (Random Access Memory)).

[0150] Although the present invention has been described above with reference to embodiments, the present invention is not limited to the above embodiments. Various modifications to the structure and details of the present invention can be made that are understandable to those skilled in the art within the scope of the present invention. That is, the present invention naturally includes the full disclosure, including the claims, and various modifications and alterations that those skilled in the art could make in accordance with the technical idea. Furthermore, each disclosure of the above-mentioned patent documents and other references is incorporated herein by reference.

[0151] 1, 1A Vehicle terminal 2, 2A Server device 3 Model execution device 5 Map information 6 User-related information 7 Model information 11, 21, 31 Communication unit 12, 22, 32 Storage unit 13, 23, 33 Input unit 14, 24, 34 Control unit 15 Sensor group 16 Display unit 17 Sound output unit

Claims

1. An information providing device comprising: an instruction information generation unit that generates instruction information including a request for recommended spots based on the current status of the moving state of a mobile body or at least one of the user's profile information and passenger information relating to the user and passengers riding in the mobile body; an acquisition unit that inputs the instruction information into a large-scale language model and acquires response information output from the large-scale language model; and a notification unit that, based on the response information, notifies the occupants of the mobile body of the spot and the reason for selecting the spot for the passengers.

2. The information providing device according to claim 1, wherein the instruction information generation unit generates instruction information including the specification of the format of the response output by the large-scale language model based on the attributes of the occupant or the movement history of the mobile body, and the notification unit notifies the occupant of the reason for the selection expressed in the response information according to the format.

3. The information providing device according to claim 2, wherein the instruction information generation unit generates instruction information specifying that the reason for selection is concise when the occupants include a child, or when the moving body is moving through a road section where the travel history is less than or equal to a predetermined number of times, and the notification unit notifies the occupants of the reason for selection expressed concisely in the response information.

4. The information providing device according to claim 1, wherein the instruction information generation unit generates instruction information that further includes content inquiring about the reputation of the spot, the acquisition unit acquires the response information that includes information regarding the reputation, and the notification unit notifies the occupant of the information regarding the reputation as one of the reasons for the selection.

5. The information providing device according to claim 1, wherein the passenger information includes information indicating the relationship between the user and the passenger.

6. The information providing device according to claim 1, wherein the passenger information includes the passenger's profile information.

7. The information providing device according to claim 1, wherein the passenger information is generated based on user input when a passenger is detected when boarding the mobile vehicle.

8. The information providing device according to claim 1, wherein the instruction information generation unit generates instruction information that further includes instructions for generating questions for selecting a spot based on at least one of the current situation or the profile information, and the acquisition unit inputs the instruction information to a large language model to acquire questions output from the large language model, and then supplies the crew's answers to the questions to the large language model to acquire response information generated by the large language model based on the answers.

9. The information providing device according to claim 8, wherein the instruction information generation unit generates instruction information including an instruction to generate a predetermined number of options to which numbers are assigned as candidates for the answer.

10. The information providing device according to claim 8, wherein the instruction information generation unit generates instruction information that includes an instruction to generate a question for confirming the current feelings of the occupant.

11. A computer-based method for providing information, comprising: an instruction information generation step of generating instruction information that includes a request for recommended spots based on at least one of the current status of the moving state of a mobile body or user profile information and passenger information relating to the user and passengers riding in the mobile body; an acquisition step of acquiring response information output from a large-scale language model by inputting the instruction information into the large-scale language model; and a notification step of notifying the occupants of the mobile body of the spots and the reasons for selecting those spots for the passengers based on the response information.

12. A program that causes a computer to function as a notification unit that generates instruction information including a request for recommended spots based on the current status of the moving state of a mobile body, or at least one of the user's profile information and passenger information relating to the user and passengers riding in the mobile body; an acquisition unit that acquires response information output from a large-scale language model by inputting the instruction information into the large-scale language model; and a notification unit that notifies the occupants of the mobile body of the spots and the reasons for selecting those spots for the passengers based on the response information.

13. A storage medium characterized by storing the program described in claim 12.

Images