Information processing device and information processing method

Abstract

Large-scale language models increase the probability of providing users with appropriate answers. [Solution] The information processing device 1 includes a question acquisition unit 131 that acquires question information indicating a user's question, a large-scale language model 3 that generates answers corresponding to the question information and is capable of referencing provided data which is data provided in a service provided by a predetermined business operator. The question information is input to the large-scale language model 3, and the large-scale language model 3 is made to refer to activity information indicating the user's activities and provided data, thereby acquiring answer information indicating the user's activities indicated by the activity information and the answers corresponding to the questions indicated by the question information, and an output unit 134 that outputs the answer information acquired by the answer acquisition unit 133.

Description

Technical Field

[0004] , , ,

[0006] , , , , , , , , , , ,

[0003] , , , , , ,

[0005]

[0001] The present invention relates to an information processing apparatus and an information processing method.

Background Art

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Patent Document 3

Summary of the Invention

Problems to be Solved by the Invention

[0004] By outputting an answer based on the user's attributes, the probability of outputting an answer corresponding to the user increases. However, it cannot be said that an answer suitable for the user is necessarily output only based on the user's attributes.

[0005] Therefore, the present invention has been made in view of these points, and an object thereof is to increase the probability that a large language model gives an appropriate answer to a user. <0W000033 <000003A It should be noted that there are some seemingly incorrect tags in the original text like <s

[0002] and <0G000024> and <0W000033>. I translated them as they are while pointing out these potential errors. If they are meant to be something else, the translation might need to be adjusted accordingly.An information processing device according to a first aspect of the present invention includes: a question acquisition unit that acquires question information indicating a user's question; an answer acquisition unit that inputs the question information into a large-scale language model that generates an answer corresponding to the question information and is capable of referencing provided data which is data provided in a service provided by a predetermined business operator, and causes the large-scale language model to refer to activity information indicating the user's activity and the provided data, thereby acquiring answer information indicating the user's activity indicated by the activity information and the answer corresponding to the question indicated by the question information; and an output unit that outputs the answer information acquired by the answer acquisition unit.

[0007] The response acquisition unit may acquire the user's activity information, generate a prompt that includes the acquired activity information and the question information, instructing the generation of a response corresponding to the user's activity indicated by the activity information and the question indicated by the question information, input the prompt to the large-scale language model, and acquire the response information from the large-scale language model.

[0008] The large-scale language model has a storage unit that stores user identification information for identifying the user and the user's activity information in association with each other, and the large-scale language model is able to refer to the user identification information and the activity information. The answer acquisition unit includes the user's user identification information and the question information, and generates a prompt that instructs the large-scale language model to refer to the activity information associated with the user identification information and generate an answer corresponding to the user's activity indicated by the activity information and the question indicated by the question information, inputs the prompt to the large-scale language model, and acquires the answer information from the large-scale language model.

[0009] The large-scale language model is capable of referencing the activity information, and the information processing device has a setting unit that sets the range of the provided data that the large-scale language model can refer to when generating an answer corresponding to the user's question information, based on the activity information, and the answer acquisition unit may, after the setting unit has made the setting, input the question information to the large-scale language model and acquire from the large-scale language model the provided data within the range set by the setting unit and answer information that indicates an answer corresponding to the question indicated by the question information.

[0010] The setting unit may set the provided data corresponding to the services the user has subscribed to as data that the large-scale language model can refer to, and if the activity information satisfies predetermined conditions, it may also set the provided data corresponding to services the user has not subscribed to as data that the large-scale language model can refer to.

[0011] The predetermined condition may be that the frequency of obtaining the question information from the user, or the frequency of outputting the answer information to the user, exceeds a predetermined threshold. The aforementioned predetermined conditions may also include the fact that the usage status of the services provided by the business operator operating the information processing device or the aforementioned predetermined business operator satisfies the predetermined usage conditions.

[0012] The response acquisition unit may, based on the activity information, identify the data to be referenced by the large-scale language model from among the provided data, and may also reference the identified data to the large-scale language model when causing the large-scale language model to generate a response corresponding to the user's question information.

[0013] The response acquisition unit may, based on the activity information, identify web data to be referenced by the large-scale language model, and may also reference the identified web data to the large-scale language model when causing the large-scale language model to generate an answer corresponding to the user's question information.

[0014] The provided data provided in the service provided by the specified business operator is provided to the user on the condition that the user is subscribed to the service, and the output unit may output the provided data used in the response indicated by the response information acquired by the response acquisition unit, or recommendation information that recommends subscribing to the service corresponding to the provided data.

[0015] The output unit may identify the service corresponding to the provided data used in the response indicated by the response information acquired by the response acquisition unit, refer to the service subscription information indicating the user's subscription status for each of the multiple services, and output the recommendation information if the identified service is a service that the user does not subscribe to.

[0016] The output unit may output the recommendation information, which includes information indicating that a benefit will be granted to the user on the condition that the user purchases the provided data or subscribes to the service.

[0017] The question acquisition unit acquires the question information from the user's terminal, and the output unit identifies the service corresponding to the provided data used in the answer indicated by the answer information acquired by the answer acquisition unit, refers to service subscription information indicating the user's subscription status for each of the multiple services, and may display the answer information on the user's terminal in different ways depending on whether the identified service is a service the user is subscribed to or a service the user is not subscribed to.

[0018] The activity information may also include information indicating at least one of the following: the location of the terminal used by the user; the user's payment history; the user's contract for the services used by the user; the user's usage history of the provided data; the user's conversation history; the user's email history; the user's usage history of social media networks; and the user's usage history of the provided data of the large-scale language model in response to the user's questions.

[0019] The information processing method according to the second aspect of the present invention includes steps of: a computer acquiring question information indicating a user's question; inputting the question information into a large language model that generates an answer corresponding to the question information and can refer to provided data which is data provided in a service provided by a predetermined business operator, and causing the large language model to refer to activity information indicating the user's activity and the provided data, thereby acquiring answer information indicating an answer corresponding to the activity of the user indicated by the activity information and the question indicated by the question information; and outputting the acquired answer information.

Advantages of the Invention

[0020] According to the present invention, there is an effect that the probability that the large language model gives an appropriate answer to the user can be increased.

Brief Description of the Drawings

[0021] [Figure 1] It is a diagram showing an overview of an information processing apparatus. [Figure 2] It is a diagram showing a functional configuration of an information processing apparatus. [Figure 3] It is a diagram showing an example of service subscription information. [Figure 4] It is a diagram showing an example of access range information. [Figure 5] It is a diagram showing an example of a database for RAG. [Figure 6] It is a diagram showing an example of user-specific activity information. [Figure 7] It is a diagram showing an example in which recommendation information is output to the user terminal 2 together with answer information. [Figure 8] It is a flowchart showing an example of a processing flow in an information processing apparatus.

Embodiments for Carrying Out the Invention

[0022] [Overview of Information Processing Apparatus 1] Figure 1 shows an overview of the information processing device 1. The information processing device 1 is a computer that provides an answer service, outputting answer information that shows the answer to a user's question using a large-scale language model. The information processing device 1 is communicated with the user terminal 2, such as a personal computer or smartphone used by the user, and the large-scale language model 3. The large-scale language model 3 supports Retrieval-Augmented Generation (RAG), and can access the provided data, which is data that shows the content provided in the content-providing service, and can generate answers corresponding to said provided data. The provided data is provided by a server different from the information processing device 1, but is not limited to this, and may be provided by the information processing device 1.

[0023] The content provision service is a service distinct from the response service provided by the information processing device 1, and is provided by one or more designated businesses. Examples of such services include a motion distribution service that distributes video data as provided data, an article distribution service that distributes newspaper article data and related data, and a gourmet service that provides gourmet information. Furthermore, the data provided in these services is provided to users on the condition that the user subscribes to the service.

[0024] Here, the information processing device 1 and the large-scale language model 3 will be described as being available to the user, for example, by the user paying a predetermined fee to the operator of the information processing device 1. However, the user may be able to use the information processing device 1 and the large-scale language model 3 without paying a predetermined fee to the operator.

[0025] The information processing device 1 obtains question information from the user terminal 2 (Figure 1, (1)). The information processing device 1 inputs the question information into the large-scale language model 3 (Figure 1, (2)).

[0026] In this embodiment, the information processing device 1 controls the large-scale language model 3 to refer to the user's activities indicated by the activity information and to output a response corresponding to the user's activities. By doing so, the information processing device 1 can increase the probability that the large-scale language model 3 will provide an appropriate response to the user.

[0027] When question information is input, the large-scale language model 3 refers to the activity information and the data (not shown) that the large-scale language model 3 has accumulated to generate answers (Figure 1, (3)), and generates answer information that shows the question indicated by the question information and the answer corresponding to the user's activity. The large-scale language model 3 then outputs the generated answer information to the information processing device 1 (Figure 1, (4)). The information processing device 1 retrieves the answer information from the large-scale language model 3 and outputs the retrieved answer information to the user terminal 2 (Figure 1, (5)).

[0028] [Functional configuration of the information processing device 1] Next, the functions of the information processing device 1 will be described. Figure 2 is a diagram showing the functional configuration of the information processing device 1 according to the first embodiment. The information processing device 1 includes a communication unit 11, a storage unit 12, and a control unit 13.

[0029] The communication unit 11 is a communication interface for sending and receiving data with external devices such as user terminals 2 and large-scale language models 3 via a communication network such as the Internet. The storage unit 12 is a storage medium for storing various types of data, and includes ROM (Read Only Memory), RAM (Random Access Memory), and hard disks. The storage unit 12 stores programs to be executed by the control unit 13. The storage unit 12 stores programs that cause the control unit 13 to function as a question acquisition unit 131, a setting unit 132, an answer acquisition unit 133, and an output unit 134.

[0030] Furthermore, the storage unit 12 stores service subscription information that indicates the user's subscription status to services provided by one or more predetermined service providers. Figure 3 shows an example of service subscription information. As shown in Figure 3, the service subscription information is information that associates, for example, a user ID as user identification information to identify a user with the service ID of the service to which the user has subscribed. As shown in Figure 3, there are multiple services provided by one or more predetermined service providers, and a user can subscribe to one or more services. For example, as shown in Figure 3, it can be confirmed that a user with user ID "U0001" has subscribed to the service with service ID "SV0001" and the service with service ID "SV0003". In addition, the services provided by the predetermined service providers include paid services that can be used by the user upon payment of a predetermined fee.

[0031] Furthermore, the storage unit 12 stores access range information for setting the available data that the large-scale language model 3 can use when generating responses based on the user's service subscription status. The storage unit 12 stores access range information that associates the user's service subscription status with data IDs, which are data identification information for identifying the data corresponding to one or more pieces of available data that correspond to that subscription status.

[0032] Figure 4 shows an example of access scope information. As shown in Figure 4, it can be seen that the access scope information is associated with a service ID, a service name, and one or more data IDs. The provided data indicated by the data ID represents a group of data provided by the service, which is data provided to the user on the service when the user subscribes to the service, or explanatory data that explains the content of such data. By setting the access scope for this provided data based on the access scope information, the large-scale language model 3 can access the provided data in a response service that is different from the service corresponding to the provided data.

[0033] Furthermore, the data provided by the data ID associated with the service ID includes not only the data provided to the user in the service indicated by that service ID, but also data provided to service subscribers in other services different from the service in question, which the user is not subscribed to. These other services may be, for example, services provided by the same company that provides the service, or services provided by other companies different from the same company.

[0034] For example, if the service is an article viewing service, the access scope information would include, in addition to text data showing newspaper articles and explanatory data showing summaries of those articles, provided to users who subscribe to the service via communication lines such as the internet, as well as explanatory data describing the content of videos provided to subscribers of video streaming services partnered with the service. This allows users to obtain answers to their questions based on data provided by other services they do not subscribe to.

[0035] Furthermore, the storage unit 12 stores a RAG database, which is a database accessible by the large-scale language model 3, that associates a service ID with a data ID of the data provided to users subscribed to the service of that service ID, and with the data group (provided data) corresponding to that data ID. Figure 5 shows an example of the RAG database. As shown in Figure 5, it can be confirmed that the service ID, data ID, and data group are associated. Note that the RAG database may be stored in another device managed by a designated service provider that provides the service. Also, provided data may be shared across multiple services, in which case multiple service IDs may be associated with a single data ID in the RAG database.

[0036] Furthermore, the storage unit 12 stores user-specific activity information that associates the user's user ID with the user's activity information. Figure 6 is a diagram showing an example of user-specific activity information. As shown in Figure 6, the activity information includes location information indicating the location of the terminal used by the user indicated by the user ID, and payment history information indicating the user's payment history. The activity information is not limited to what is shown in Figure 6, but may also include information indicating at least one of the following: the user's contract with the service provider related to the service the user is using, the usage history of data provided by the service, the user's conversation history with the large-scale language model 3, the user's email history, the user's usage history of the social media network, and the usage history of data provided by the large-scale language model 3 in response to the user's questions to the information processing device 1's answer service. The activity information includes the activity time, which is the time when the user performed the activity. This allows the information processing device 1 to obtain the user's activity information for a predetermined period. In this embodiment, the storage unit 12 stores the user-specific activity information, but it is not limited to this, and the storage unit of another device that the large-scale language model 3 can access may store the user-specific activity information.

[0037] The control unit 13 is, for example, a CPU (Central Processing Unit). The control unit 13 functions as a question acquisition unit 131, a setting unit 132, an answer acquisition unit 133, and an output unit 134 by executing a program stored in the memory unit 12.

[0038] The question acquisition unit 131 acquires question information indicating the user's question. For example, the question acquisition unit 131 acquires question information from the user terminal 2. In this case, the question acquisition unit 131 causes the user terminal 2 to display a reception screen for receiving questions from the user. Then, the question acquisition unit 131 acquires a prompt as question information indicating the user's question entered via the reception screen, and the user's user ID from the user terminal 2. The question acquisition unit 131 may acquire the user's user ID in advance before receiving the question information.

[0039] The question acquisition unit 131 acquired a prompt as question information entered by the user from the user terminal 2, but is not limited to this. The question acquisition unit 131 may also acquire question information indicating the user's question via a server (not shown) that provides services offered by a designated business operator.

[0040] For example, if the service provided by a designated business operator is a video distribution service that distributes videos, the server displays a page on the user terminal 2 that provides the video distribution service and includes an input field for receiving questions from the user. The server then transmits question information indicating the question obtained from the user terminal 2 to the information processing device 1. The question acquisition unit 131 acquires the question information indicating the user's question from the server.

[0041] Furthermore, a generation device (not shown) for receiving questions from the user is provided, and the generation device may generate question information by changing the expression of the question received from the user terminal 2 to content suitable as a prompt to be input to the large-scale language model 3. Then, the question acquisition unit 131 may acquire the question information from the generation device.

[0042] The configuration unit 132 refers to the service subscription information stored in the storage unit 12 to identify the service subscription status of the user who asked the question. For example, in response to the question acquisition unit 131 acquiring the question information, the configuration unit 132 refers to the service subscription information stored in the storage unit 12 and identifies the service ID associated with the user ID of the user who asked the question, which was acquired by the question acquisition unit 131, thereby identifying the user's service subscription status. If the user is subscribed to multiple services, the configuration unit 132 identifies the service ID for each of the multiple services associated with the user ID of the user who asked the question.

[0043] Furthermore, the configuration unit 132 identifies one or more data IDs stored in the storage unit 12 in association with the identified subscription status. If a user is subscribed to multiple services, the configuration unit 132 identifies multiple data IDs based on the user's subscription status for each of the multiple services. For example, the configuration unit 132 refers to the access range information stored in the storage unit 12 and identifies one or more data IDs associated with each of the identified one or more service IDs.

[0044] The configuration unit 132 identifies the service subscription status of the user who asked the question in response to the question acquisition unit 131 acquiring the question information, but is not limited to this. If the configuration unit 132 has already acquired the user's user ID before the question acquisition unit 131 acquires the question information, it may also identify the service subscription status of the user who asked the question in response to the question acquisition unit 131 acquiring the user ID.

[0045] When the large-scale language model 3 generates an answer corresponding to the question information in response to the identification of one or more data IDs, the setting unit 132 configures the large-scale language model 3 so that the provided data indicated by each of the identified data IDs is available to the large-scale language model 3.

[0046] For example, when the configuration unit 132 identifies one or more data IDs corresponding to a user ID, it associates the user ID with the data ID and stores it in the storage unit 12 as access rights information that the large-scale language model 3 can refer to, thereby configuring the large-scale language model 3. As a result, the large-scale language model 3 can refer to the access rights information, access the provided data indicated by the data ID associated with the user ID of the user who asked the question, from among the data stored in the RAG database, and generate answer information based on said provided data.

[0047] The response acquisition unit 133, after the settings for accessing the data provided by the large-scale language model 3 have been configured by the setting unit 132, inputs question information into the large-scale language model 3 and causes the large-scale language model 3 to refer to the provided data, thereby acquiring response information that indicates the user's activities indicated by the user's activity information and the answers to the questions indicated by the question information.

[0048] Specifically, first, the answer acquisition unit 133 refers to the user-specific activity information stored in the storage unit 12 and acquires the user activity information associated with the user ID acquired by the question acquisition unit 131. The answer acquisition unit 133 may also acquire activity information for a predetermined period from the current time to a predetermined time prior. The answer acquisition unit 133 includes the acquired activity information and question information and generates a prompt that instructs the generation of an answer corresponding to the user activity indicated by the activity information and the question indicated by the question information.

[0049] For example, a prompt aligns with the user's activities indicated by the activity information, instructs the system to provide an answer to a question, and includes instructions that cause the large-scale language model 3 to identify the data ID associated with the user ID in the access rights information. The prompt also includes instructions that cause the system to generate answer information by referencing pre-stored data and data associated with the identified data ID in the RAG database, and instructions that, if the data used in the answer is provided data, generate answer information that includes the data ID of the provided data.

[0050] The response acquisition unit 133 inputs the generated prompt to the large-scale language model 3 and acquires response information from the large-scale language model 3. If the large-scale language model 3 generates a response corresponding to the provided data, the response acquisition unit 133 acquires the response information along with a data ID indicating the provided data.

[0051] The response acquisition unit 133 may, based on the user's activity information, identify the data to be referenced by the large-scale language model 3 from among the provided data. For example, the response acquisition unit 133 may, based on the user's activity information and the question indicated by the question information, identify the user's activity area and the user's attributes corresponding to the question, and then identify the provided data corresponding to the identified activity area and attributes. In this case, it is assumed that each piece of provided data, or each data group constituting the provided data, has associated data attributes.

[0052] For example, if the question information indicates "Can you recommend a good restaurant?", the attribute corresponding to the question is identified as "gourmet". Then, the answer acquisition unit 133 identifies the attribute corresponding to the user's gourmet tastes and attributes related to gourmet tastes based on the user's activity information. For example, the answer acquisition unit 133 identifies the restaurants the user has visited and the user's residential area based on the user's payment history information and terminal location information. Then, the answer acquisition unit 133 identifies the attribute corresponding to the restaurants the user is likely to visit as the gourmet genre, which is the attribute corresponding to the user's gourmet tastes, and identifies the user's activity area as an attribute related to the user's gourmet tastes.

[0053] For example, suppose the response acquisition unit 133 identifies the gourmet genre corresponding to the user as "Chinese cuisine" and the user's activity area as "Shibuya". The response acquisition unit 133 identifies the data or data set that corresponds to the gourmet service and has attributes corresponding to both the area "Shibuya" and the gourmet genre "Chinese cuisine" as the data to be referenced by the large-scale language model 3.

[0054] The answer acquisition unit 133 then causes the large-scale language model 3 to generate an answer corresponding to the user's question information, by having the large-scale language model 3 refer to the identified provided data. The answer acquisition unit 133 may also include a prompt that indicates the identified attribute and instructs the large-scale language model 3 to generate an answer by referring to the provided data corresponding to that attribute, thereby causing the large-scale language model 3 to refer to the identified provided data.

[0055] Furthermore, the response acquisition unit 133 may identify web data to be referenced by the large-scale language model 3 based on activity information, and may also reference the identified web data to the large-scale language model 3 when causing the large-scale language model 3 to generate an answer corresponding to the user's question information.

[0056] For example, as described above, suppose the response acquisition unit 133 identifies the gourmet genre corresponding to the user as "Chinese cuisine" and the user's activity area as "Shibuya". In this case, the response acquisition unit 133 may include in its prompts an instruction that indicates the identified attributes and instructs the generation of a response by referring to data corresponding to those attributes, and an instruction that the data to be referenced may include the address of a website that provides gourmet information publicly available on the web, and that instructs the generation of a response by referring to that website, and cause the large-scale language model 3 to refer to this data. By doing so, the information processing device 1 can increase the probability that the large-scale language model 3 will generate a response corresponding to the user's attributes.

[0057] Furthermore, the response acquisition unit 133 generates a prompt that includes the acquired activity information and instructs the generation of responses corresponding to the user's activities indicated by the activity information and the questions indicated by the question information, but is not limited to this. For example, the large-scale language model 3 may be instructed to refer to the user's activity information included in the user-specific activity information and generate a prompt that instructs the generation of responses corresponding to the user's activities indicated by the activity information and the questions indicated by the question information.

[0058] In this case, the large-scale language model 3 is capable of referencing user-specific activity information that associates the user ID with the activity information. The answer acquisition unit 133 then generates a prompt that includes the user's user ID and question information, instructing the large-scale language model 3 to refer to the activity information associated with the user ID and generate answers corresponding to the user's activities indicated by the activity information and the questions indicated by the question information. The answer acquisition unit 133 then inputs the generated prompt to the large-scale language model 3 and obtains the answer information from the large-scale language model 3. In this way, the information processing device 1 can cause the large-scale language model 3 to generate answer information corresponding to the user's activity information, similar to when generating a prompt that includes the user's activity information.

[0059] Furthermore, the configuration unit 132 may set the range of provided data that the large-scale language model 3 can access when generating answers corresponding to the user's question information, based on the activity information. For example, the configuration unit 132 identifies the user's activity area and the user's attributes corresponding to the question based on the user's activity information and the question indicated by the question information, and sets the provided data corresponding to the identified activity area and attributes so that the large-scale language model 3 can access it. For example, the configuration unit 132 sets the provided data so that the large-scale language model 3 can access it by setting access rights to the provided data corresponding to the identified activity area and attributes for the large-scale language model 3.

[0060] Here, the setting unit 132 may not only set the provided data corresponding to the services the user has subscribed to as data that the large-scale language model 3 can refer to, but may also set the provided data corresponding to services the user has not subscribed to as data that the large-scale language model 3 can refer to if the activity information satisfies predetermined conditions.

[0061] For example, the predetermined conditions include the frequency of obtaining question information from the user or the frequency of outputting answer information to the user exceeding a predetermined threshold, or the usage status of a service that provides content to the user, provided by the operator of the information processing device 1 or a predetermined operator, meeting predetermined usage conditions. Furthermore, the predetermined usage conditions include, for example, the user using a service designated in advance by the operator of the information processing device 1 or a predetermined operator, or the user using such a service for a predetermined period. For example, the setting unit 132 may set the provided data corresponding to services not subscribed to by the user as data that the large-scale language model 3 can reference, on the condition that the user is subscribed to a predetermined communication plan offered by a predetermined telecommunications carrier. In this case, the output unit 134, described later, may output the predetermined conditions and information indicating that the amount of provided data that the large-scale language model 3 can reference increases when the predetermined conditions are met, along with the answer information.

[0062] Furthermore, the response acquisition unit 133 may input question information to the large-scale language model 3 after the setting unit 132 has set the range of activity information that can be referenced, and acquire from the large-scale language model 3 the provided data within the range set by the setting unit 132 and response information that indicates the answer corresponding to the question indicated by the question information. In this way, the information processing device 1 can increase the probability that the large-scale language model 3 will provide an appropriate answer to the user by allowing the large-scale language model 3 to access activity information from the user's activity information that increases the probability that the response information corresponds to the user when generating response information.

[0063] The output unit 134 outputs the answer information acquired by the answer acquisition unit 133. The output unit 134 outputs the answer information acquired by the answer acquisition unit 133 to the source from which the question information was acquired by the question acquisition unit 131. For example, if the question acquisition unit 131 acquires question information from the user terminal 2, the output unit 134 outputs the answer information to the user terminal 2. Also, if the question acquisition unit 131 acquires question information from a server that provides services provided by a predetermined business operator, the output unit 134 outputs the answer information acquired by the answer acquisition unit 133 to that server.

[0064] Here, the output unit 134 may identify the service corresponding to the provided data used in the response indicated by the response information acquired by the response acquisition unit 133. The output unit 134 may then refer to service subscription information indicating the user's subscription status for each of the multiple services, and display the response information on the user terminal 2 in different ways depending on whether the identified service is a service the user is subscribed to or a service the user is not subscribed to.

[0065] The response information corresponds to the provided data when, for example, the response acquisition unit 133 acquires the response information including the data ID of the provided data. The output unit 134 determines that the response information corresponds to the provided data when the response acquisition unit 133 acquires the data ID of the provided data. The output unit 134 then refers to the RAG database and identifies the service corresponding to the provided data used in the response by identifying the service ID associated with the data ID indicating the provided data corresponding to the response information.

[0066] Furthermore, the output unit 134 refers to the service subscription information and determines that the user is not subscribed to the service corresponding to the provided data if the identified service ID is not associated with the user ID of the user who made the question.

[0067] If the identified service is not a service the user subscribes to, the output unit 134 changes the display format of the response information to emphasize that the response information was generated based on data provided by a service the user does not subscribe to, and displays the response information on the user terminal 2. On the other hand, if the identified service is not a service the user subscribes to, the output unit 134 displays the response information on the user terminal 2 without changing the display format of the response information.

[0068] Furthermore, the output unit 134 may output recommendation information if the service corresponding to the provided data indicated by the response information is a service that the user has not subscribed to. In this case, when the output unit 134 identifies a service corresponding to the provided data indicated by the response information, and the identified service is a service that the user has not subscribed to, it generates recommendation information and outputs the recommendation information. For example, the output unit 134 generates recommendation information that recommends a service that the user has not subscribed to, provided by a service provider that provides the service corresponding to the provided data, and outputs the recommendation information.

[0069] Here, the output unit 134 may generate recommendation information that shows links to websites for purchasing content or subscribing to a service that provides the provided data. For example, link information that associates a link to a page for purchasing content indicated by the provided data or subscribing to a service corresponding to the provided data with the data ID of the provided data is stored in the storage unit 12. Here, in order to identify that a user has subscribed to a service based on the recommendation information, the output unit 134 may include identification information for identifying the recommendation information as a parameter in the link.

[0070] Then, if the response information corresponds to the provided data, the output unit 134 generates response information including the recommendation information by embedding the recommendation information stored in the storage unit 12, associated with the data ID of the provided data, into the response information. In this way, the information processing device 1 can quickly allow users to purchase content or subscribe to services from the page indicated by the link if they wish to do so.

[0071] Furthermore, the output unit 134 may generate recommendation information that includes benefit information indicating that a benefit will be granted to the user on the condition that the user purchases the provided data or subscribes to a service corresponding to the provided data, if the response information corresponds to the provided data. Here, the benefit may be, for example, a discount that can be used when subscribing to or using a service corresponding to the provided data or a recommended service, or the granting of points that can be used in these services.

[0072] In this case, the output unit 134 may determine the content of the service benefits based on the user's usage of the large-scale language model 3. For example, the output unit 134 may obtain usage frequency information from the large-scale language model 3 indicating how often the user uses the large-scale language model 3, and determine the content of the benefits so that the more frequently the user uses it, the higher the value of the benefits, or the more frequently the questions and answers correspond to the service, the higher the value of the benefits. The output unit 134 may then generate recommendation information that includes benefit information indicating the determined benefits. By doing so, the information processing device 1 can increase the probability that the user will subscribe to the service compared to when no benefits are provided.

[0073] The output unit 134 then outputs the answer information and, if recommendation information is generated by the output unit 134, outputs the recommendation information to the source of the question information. For example, the output unit 134 outputs the recommendation information to the source of the question information by outputting the answer information that includes the recommendation information.

[0074] Figure 7 shows an example where recommendation information is output to the user terminal 2 along with the response information. In the example shown in Figure 7, it can be seen that information indicating that the response information was generated based on the provision data of a service that the user has not subscribed to is displayed. Furthermore, in the example shown in Figure 7, recommendation information that recommends a service corresponding to the response information is displayed, and it can be seen that this recommendation information includes a link to the service. As a result, the information processing device 1 can increase the probability that the user will subscribe to the service corresponding to the response information.

[0075] [flowchart] Next, the processing flow of the information processing device 1 will be explained. Figure 8 is a flowchart showing an example of the processing flow in the information processing device 1. Figure 8 explains the processing flow from when the information processing device 1 acquires question information to when it outputs answer information. It is assumed that the service subscription information and access range information are stored in the storage unit 12 before the flowchart shown in Figure 8 is executed.

[0076] First, the question acquisition unit 131 acquires question information indicating the user's question and the user's user ID (S1). Next, the setting unit 132 refers to the service subscription information stored in the storage unit 12 and identifies the service ID associated with the user ID of the user who asked the question, thereby determining the user's service subscription status (S2).

[0077] Next, the configuration unit 132 refers to the access range information stored in the storage unit 12 and identifies one or more data IDs associated with the identified service ID, thereby identifying the data ID associated with the user's service subscription status (S3).

[0078] Next, the setting unit 132 configures the large-scale language model 3 so that the data indicated by each of the one or more data IDs identified by the setting unit 132 is available when the large-scale language model 3 generates an answer corresponding to the question information (S4).

[0079] Next, the response acquisition unit 133 refers to the user-specific activity information stored in the storage unit 12 and acquires the user activity information associated with the user ID of the user who asked the question (S5). The response acquisition unit 133 acquires the response information corresponding to the acquired activity information and question information, and the data ID of the data corresponding to the response from the large-scale language model 3 (S6).

[0080] Specifically, the response acquisition unit 133 generates a prompt that includes the acquired activity information and question information, instructing the generation of an answer corresponding to the user's activity indicated by the activity information and the question indicated by the question information. By inputting this prompt into the large-scale language model 3, the large-scale language model 3 obtains from the large-scale language model 3 the response information indicating the answer corresponding to the activity information and the question indicated by the question information, as well as the data ID of the data used in the answer. Subsequently, the output unit 134 outputs the response information acquired by the response acquisition unit 133 to the source of the question information acquired by the question acquisition unit 131 (S7).

[0081] [Differentiation] In the above-described embodiment, when the information processing device 1 generates an answer corresponding to the question information, the data indicated by one or more data IDs associated with the user's service subscription status is made available to the large-scale language model 3. However, the invention is not limited to this. The large-scale language model 3 may be configured to make predetermined provided data available regardless of the user's service subscription status.

[0082] In this case, if the provided data is used in the response indicated by the response information obtained by the response acquisition unit 133, the output unit 134 may generate recommendation information recommending that the user subscribe to the provided data or the service corresponding to the provided data, regardless of whether the user is subscribed to the service corresponding to the provided data, and output the generated recommendation information.

[0083] Furthermore, in the above-described embodiment, the large-scale language model 3 is provided separately from the information processing device 1, but this is not limited to this configuration; the information processing device 1 may also function as the large-scale language model 3.

[0084] [Effects of Information Processing Device 1] As described above, the information processing device 1 according to this embodiment is a large-scale language model 3 that acquires question information indicating a user's question and generates an answer corresponding to the question information. The information processing device 1 inputs the question information into the large-scale language model 3, which is data provided in a service provided by a predetermined business operator, and causes the large-scale language model 3 to refer to activity information indicating the user's activity and the provided data. In doing so, it acquires answer information indicating the user's activity indicated by the activity information and the answer corresponding to the question indicated by the question information, and outputs the acquired answer information. In this way, the information processing device 1 can increase the probability that the large-scale language model 3 will provide an appropriate answer to the user.

[0085] Furthermore, this invention will make it possible to contribute to Goal 9 of the United Nations-led Sustainable Development Goals (SDGs), "Build resilient infrastructure, promote inclusive and sustainable industrialization and foster innovation."

[0086] Although the present invention has been described above using embodiments, the technical scope of the present invention is not limited to the scope described in the above embodiments, and various modifications and changes are possible within the scope of its gist. For example, all or part of the apparatus can be configured by functionally or physically distributing and integrating in any unit. Furthermore, new embodiments resulting from any combination of multiple embodiments are also included in the embodiments of the present invention. The effects of the new embodiments resulting from the combinations are combined with the effects of the original embodiments. [Explanation of symbols]

[0087] 1. Information Processing Device 2 User terminals 11 Communications Department 12 Storage section 13 Control Unit 131 Question acquisition part 132 Settings Section 133 Answer acquisition part 134 Output section

Claims

1. A question acquisition unit that acquires question information indicating the user's question, A large-scale language model that generates answers corresponding to the aforementioned question information, wherein the large-scale language model, which can reference provided data which is data provided in a service provided by a predetermined business operator, sets the range of the provided data that can be referenced when generating answers corresponding to the user's question information, based on activity information indicating the user's activities, and the setting unit sets the range of provided data corresponding to services that the user has not subscribed to as data that the large-scale language model can reference when the activity information satisfies predetermined conditions, After the settings are made by the setting unit, the answer acquisition unit inputs the question information into the large-scale language model and causes it to refer to the provided data within the range set by the setting unit, thereby acquiring answer information that indicates the answer corresponding to the provided data within the range set by the setting unit and the question indicated by the question information. The output unit outputs the response information acquired by the response acquisition unit, An information processing device having

2. The response acquisition unit, after the setting unit has performed the setting, inputs the question information into the large-scale language model and causes it to refer to the activity information and the provided data within the range set by the setting unit, thereby acquiring response information that indicates the answers corresponding to the user's activities indicated by the activity information, the provided data within the range set by the setting unit, and the questions indicated by the question information. The information processing apparatus according to claim 1.

3. The response acquisition unit acquires the user's activity information, generates a prompt that includes the acquired activity information and the question information, instructs the generation of a response corresponding to the user's activity indicated by the activity information and the question indicated by the question information, inputs the prompt to the large-scale language model, and acquires the response information from the large-scale language model. The information processing apparatus according to claim 1.

4. The storage unit stores user identification information for identifying the user and the user's activity information in association with each other. The large-scale language model is capable of referencing the user identification information and the activity information, The response acquisition unit includes the user identification information of the user and the question information, generates a prompt that instructs the large-scale language model to refer to the activity information associated with the user identification information and generate a response corresponding to the user's activity indicated by the activity information and the question indicated by the question information, inputs the prompt to the large-scale language model, and acquires the response information from the large-scale language model. The information processing apparatus according to claim 1.

5. The setting unit sets the provided data corresponding to the services the user has subscribed to as data that the large-scale language model can refer to, and, if the activity information satisfies the predetermined conditions, sets the provided data corresponding to services the user has not subscribed to as data that the large-scale language model can refer to. The information processing apparatus according to claim 1.

6. The predetermined condition is that the frequency of obtaining the question information from the user, or the frequency of outputting the answer information to the user, exceeds a predetermined threshold. The information processing apparatus according to claim 1.

7. The aforementioned predetermined condition is that the usage status of the services provided by the business operator operating the information processing device or the aforementioned predetermined business operator satisfies the predetermined usage conditions. The information processing apparatus according to claim 1.

8. The response acquisition unit, based on the activity information, identifies the data to be referenced by the large-scale language model from the provided data, and, when causing the large-scale language model to generate a response corresponding to the user's question information, it references the identified data to the large-scale language model. The information processing apparatus according to claim 1.

9. The response acquisition unit identifies web data to be referenced by the large-scale language model based on the activity information, and, when causing the large-scale language model to generate a response corresponding to the user's question information, it references the identified web data to the large-scale language model. The information processing apparatus according to claim 1.

10. The activity information includes at least one of the following: the location of the terminal used by the user; the user's payment history; the user's contract for the services used by the user; the usage history of the provided data; the user's conversation history; the user's email history; the user's social media network usage history; and the usage history of the provided data of the large-scale language model in response to the user's questions. The information processing apparatus according to claim 1.

11. A question acquisition unit that acquires question information indicating the user's question, A large-scale language model that generates answers corresponding to the aforementioned question information, the large-scale language model which can reference provided data which is data provided in a service provided by a predetermined business operator, is input with the question information and referenced with the provided data, thereby obtaining answer information that indicates an answer corresponding to the question indicated by the question information, The output unit outputs the response information acquired by the response acquisition unit, It has, The data provided in the service provided by the aforementioned designated business operator is provided to the user on the condition that the user is subscribed to the service. The output unit identifies a service corresponding to the provided data used in the response indicated by the response information acquired by the response acquisition unit, refers to service subscription information indicating the user's subscription status for each of the multiple services, and if the identified service is a service that the user has not subscribed to, outputs information indicating that the response indicated by the response information acquired by the response acquisition unit was generated based on the provided data, or recommendation information recommending that the user subscribe to the service corresponding to the provided data. Information processing device.

12. The output unit outputs the recommendation information, which includes information indicating that a benefit will be granted to the user on the condition that the user purchases the provided data or subscribes to the service. The information processing apparatus according to claim 11.

13. A question acquisition unit that acquires question information indicating the user's question from the user's terminal, A large-scale language model that generates answers corresponding to the aforementioned question information, the large-scale language model which can reference provided data which is data provided in a service provided by a predetermined business operator, is input with the question information and referenced with the provided data, thereby obtaining answer information that indicates an answer corresponding to the question indicated by the question information, The output unit outputs the response information acquired by the response acquisition unit, It has, The output unit identifies the service corresponding to the provided data used in the response indicated by the response information acquired by the response acquisition unit, refers to service subscription information indicating the user's subscription status for each of the multiple services, and displays the response information on the user's terminal in different ways depending on whether the identified service is a service the user is subscribed to or a service the user is not subscribed to. Information processing device

14. The answer acquisition unit inputs the question information into the large-scale language model and causes the large-scale language model to refer to the activity information indicating the user's activity and the provided data, thereby acquiring answer information indicating the user's activity indicated by the activity information, the provided data, and the question indicated by the question information. The information processing apparatus according to claim 11 or 13.

15. A computer executes Steps include obtaining question information that indicates the user's question, A large-scale language model that generates answers corresponding to the question information, wherein the large-scale language model, which can reference provided data which is data provided in a service provided by a predetermined business operator, sets the range of provided data that can be referenced when generating answers corresponding to the user's question information, based on activity information indicating the user's activities, wherein if the activity information satisfies predetermined conditions, the large-scale language model sets provided data corresponding to services that the user has not subscribed to as data that can be referenced by the large-scale language model. After the above settings are made, the large-scale language model is given the question information to input and the provided data that has been set, thereby obtaining answer information that shows the answers corresponding to the provided data within the set range and the questions indicated by the question information. The steps include outputting the acquired response information, An information processing method having

16. A computer executes Steps include obtaining question information that indicates the user's question, A large-scale language model that generates answers corresponding to the aforementioned question information, the large-scale language model which can reference provided data which is data provided in a service provided by a predetermined business operator, is input with the question information and made to reference the provided data, thereby obtaining answer information that indicates an answer corresponding to the question indicated by the question information. The steps include outputting the acquired response information, It has, The data provided in the service provided by the aforementioned designated business operator is provided to the user on the condition that the user is subscribed to the service. In the output step, the computer identifies the service corresponding to the provided data used in the answer indicated by the acquired answer information, refers to service subscription information indicating the user's subscription status for each of the multiple services, and if the identified service is a service that the user does not subscribe to, outputs information indicating that the answer indicated by the acquired answer information was generated based on the provided data, or recommendation information recommending that the user subscribe to the service corresponding to the provided data. Information processing methods.

17. A computer executes The steps include obtaining question information indicating the user's question from the user's terminal, A large-scale language model that generates answers corresponding to the aforementioned question information, the large-scale language model which can reference provided data which is data provided in a service provided by a predetermined business operator, is input with the question information and made to reference the provided data, thereby obtaining answer information that indicates an answer corresponding to the question indicated by the question information. The steps include outputting the acquired response information, It has, In the output step, the computer identifies the service corresponding to the provided data used in the answer indicated by the acquired answer information, refers to service subscription information indicating the user's subscription status for each of the multiple services, and displays the answer information on the user's terminal in different ways depending on whether the identified service is a service the user is subscribed to or a service the user is not subscribed to. Information processing methods.

Images