Apparatus, method, and program for generating target behavior promotion messages

Abstract

This involves automatically generating highly effective messages that encourage action by identifying information that is likely to prompt specific actions based on an individual's circumstances. [Solution] The target behavior promotion message generation device of the present invention comprises: a first object acquisition unit that acquires object type information indicating the type of object to be provided to the user in order to promote the behavior, based on the content of the behavior to be encouraged to the user and user information about the user; a second object acquisition unit that acquires object information indicating the content of the object, based on user information and object type information, from a pre-prepared object database; and a message generation unit that generates a message to encourage the user to take the behavior using a pre-prepared language model, based on the content of the behavior, user information, and object information.

Description

Technical Field

[0001] The present invention relates to a target behavior promotion message generation device, method, and program that are implemented by a computer and generate a message for promoting a user's behavior based on an input.

Background Art

[0002] Conventionally, in the healthcare field and the like, sending messages for the purpose of making people healthier has been carried out, for example, as health guidance by nurses and the like.

[0003] In Non-Patent Document 1, which is a past study on messages in this type of health guidance, it is shown that messages should be appropriately changed according to the attributes of the target person and the user's situation regarding healthy behavior (hereinafter referred to as the "behavior change stage").

[0004] In addition, in Non-Patent Document 2, with the spread of large language models, research has been conducted to generate such messages by a computer using a technology that enables text generation based on human instructions.

[0005] In this research, health-related inspiration messages written on microblogs such as X (registered trademark) (formerly Twitter (registered trademark)) are analyzed, the beginning part of the message considered to be effective is obtained, and the obtained beginning part is used as a prompt for input to a large language model, and research is being conducted on a technology for automatically generating a health inspiration message that follows the beginning part.

Prior Art Documents

Non-Patent Documents

[0006]

Non-Patent Document 1

[0007] However, conventional technologies have the following problems.

[0008] The research disclosed in Non-Patent Document 1 discusses the need to personalize messages that promote healthy behaviors by taking into account individual attributes and circumstances. However, creating such individually tailored messages is extremely time-consuming. Furthermore, methods for automatically generating them using computers based on rules and other factors have not been revealed.

[0009] Furthermore, the research disclosed in Non-Patent Document 2 does not address message personalization.

[0010] This invention has been made in view of these circumstances, and aims to provide a target behavior promotion message generation device, method, and program that can automatically generate highly effective messages to encourage action by identifying information that makes it easier for individuals to take specific actions, according to their individual circumstances. [Means for solving the problem]

[0011] A first aspect of the present invention for achieving the above objective is a target behavior promotion message generation device comprising: a first object acquisition unit that acquires object type information indicating the type of object to be provided to the user in order to promote the behavior, based on the content of the behavior to be encouraged to the user and user information about the user, from a pre-prepared object information database; a second object acquisition unit that acquires object information indicating the content of the object, based on user information and object type information, from a pre-prepared object database; and a message generation unit that generates a message to encourage the user to take the behavior using a pre-prepared language model, based on the content of the behavior, user information, and object information.

[0012] A second aspect of the present invention is a target action promotion message generation device according to the first aspect, wherein the object information database stores the content of an action, user information, and object type information corresponding to a combination of the content of the action and the user information in association with each other.

[0013] A third aspect of the present invention is a target action promotion message generation device according to the first aspect, wherein the object database stores user information, object type information, and object information indicating the content of an object determined by a combination of user information and object type information, in an associated manner.

[0014] A fourth aspect of the present invention is a target behavior promotion message generation device of the first aspect, wherein the behavior is a behavior aimed at health.

[0015] A fifth aspect of the present invention is a target behavior facilitator message generator of the first aspect, wherein the language model is trained to output text information in response to a specified prompt, and the message includes text information.

[0016] The sixth aspect of the present invention is a method for generating a target behavior promotion message, in which a processor obtains object type information indicating the type of object to be provided to a user for promoting the behavior from a pre-prepared object information database based on the content of the behavior to be prompted to the user and user information regarding the user, obtains object information indicating the content of the object from a pre-prepared object database based on the user information and the object type information, and generates a message for inspiring the promotion of the behavior to the user using a pre-prepared language model based on the content of the behavior, the user information, and the object information. This is a target behavior promotion message generation method.

[0017] The seventh aspect of the present invention is a program for causing a computer to function as each part included in the target behavior promotion message generation device according to any one of the first to fifth aspects.

Advantages of the Invention

[0018] According to the target behavior promotion message generation device, method, and program of the present invention, for example, according to an individual's health status, by specifying information that makes it easy to perform specific health behaviors, a message with a high effect of inspiring the promotion of health behaviors can be automatically generated.

Brief Description of the Drawings

[0019] [Figure 1] FIG. 1 is a block diagram showing an example of the functional configuration of a target behavior promotion message generation device according to an embodiment of the present invention. [Figure 2] FIG. 2 is a data structure diagram showing an example of an object information database. [Figure 3] FIG. 3 is a data structure diagram showing an example of an object database. [Figure 4] FIG. 4 is a diagram showing an example of a prompt when querying a large language model. [Figure 5] FIG. 5 is a schematic diagram showing a general hardware configuration of a computer. [Figure 6] FIG. 6 is a flowchart showing an operation example performed by the target behavior promotion message generation device according to the present embodiment.

Embodiment for Carrying Out the Invention

[0020] Hereinafter, embodiments of the present invention will be described with reference to the drawings. The drawings are schematic or conceptual, and the relationship between the thickness and width of each part, the ratio of the sizes between parts, etc. are not necessarily the same as those in reality. Also, even when representing the same part, the dimensions and ratios may be represented differently depending on the drawings. In the present specification and each figure, elements that are the same as those described in the previously shown figures are given the same reference numerals, and detailed descriptions and duplicate descriptions are omitted as appropriate.

[0021] FIG. 1 is a block diagram showing an example of the functional configuration of a target behavior promotion message generation device according to an embodiment of the present invention.

[0022] The target behavior promotion message generation device 10 illustrated in FIG. 1 is a device to which the target behavior promotion message generation method according to an embodiment of the present invention is applied, and by specifying information that makes it easy to perform specific actions according to an individual's situation, it automatically generates a message with a high effect of inspiring the promotion of behavior. The individual's situation can be, for example, the individual's health status. In this case, the information that makes it easy to perform specific actions is information that makes it easy to perform specific health actions, and the message with a high effect of inspiring the promotion of behavior is a message with a high effect of inspiring the promotion of health actions.

[0023] To achieve this, the target behavior promotion message generation device 10 includes a message request reception unit 20, a message content management unit 30, a first object acquisition unit 40, an object information storage unit 41, a second object acquisition unit 50, an object storage unit 51, a message generation unit 60, a large language model storage unit 61, and a message return unit 70. The message request reception unit 20 and the message return unit 70 can communicate with an external client terminal 90.

[0024] A person who needs to generate a message can input the action they want the user to take (a), user information (b), and environmental information (c), such as the weather and the time to send the message, directly to the client terminal 90 or through software that requires message generation. The person who needs to generate the message may be the user themselves, or they may be someone other than the user, such as a service provider, doctor, or public health nurse.

[0025] The client terminal 90 can be implemented, for example, by a smartphone with a dedicated application installed. When the client terminal 90 receives input of an action to be encouraged from the user (a), user information (b) which is information about the user, and environmental information (c), it transmits the input action (a), user information (b), and environmental information (c) to the target action promotion message generation device 10.

[0026] The input action content a, user information b, and environment information c transmitted by the client terminal 90 are received by the message request receiving unit 20. The message request receiving unit 20 outputs the received action content a, user information b, and environment information c to the message content management unit 30.

[0027] The message content management unit 30 receives the action content a, user information b, and environment information c output by the message request receiving unit 20. It then outputs the action content a, user information b, and environment information c to the first object acquisition unit 40.

[0028] The first object acquisition unit 40 receives the action content a, user information b, and environment information c output from the message content management unit 30. Then, based on the action content a, user information b, and environment information c, it searches the object information database 41A stored in the object information storage unit 41 and obtains object type information d as a search result from the object information database 41A, which indicates the type of object that should be provided to the user to facilitate the action.

[0029] Figure 2 is a data structure diagram showing an example of an object information database.

[0030] The object information database 41A stores and associates action content a, user information b, environmental information c, and object type information d corresponding to combinations of action content a, user information b, and environmental information c. Action content a may include, for example, exercise promotion and smoking cessation, user information b may include user states such as interest stage, preparation stage, maintenance stage, and execution stage (also referred to as "behavioral change stages"), and environmental information c may include weather and time zones for specifying the time to send a message. Object type information d may further include object type information (major categories) d1 such as videos, spots, and hospitals, and object type information (minor categories) d2 such as the effects of exercise, stretching methods, walking courses, negative effects of smoking cessation, smoking cessation methods, and smoking cessation clinics. In the example shown in Figure 2, behavioral change stages are taken as user information b, but if user attributes that affect the content to be conveyed (gender and age) or the individual's time discount rate are known, appropriate object type information d may be added accordingly. The individual's time discount rate can be text that represents the user's characteristics, such as a self-introduction.

[0031] The first object acquisition unit 40 searches the object information database 41A based on the action content a, user information b, and environment information c, and acquires object type information d corresponding to the action content a, user information b, and environment information c as a search result, and outputs it to the message content management unit 30.

[0032] Furthermore, the method by which the first object acquisition unit 40 acquires object type information d is not limited to using the object information database 41A, but may also be one that does not use the object information database 41A.

[0033] For example, the first object acquisition unit 40 may use a large-scale language model 61A (described later) instead of an object information database 41A to acquire object type information d, and acquire object type information d by querying the large-scale language model 61A based on the action content a, user information b, and environment information c to determine what kind of objects are likely to trigger an action.

[0034] The message content management unit 30 receives object type information d output from the first object acquisition unit 40, and outputs the received object type information d, along with user information b and environment information c, to the second object acquisition unit 50.

[0035] The second object acquisition unit 50 receives object type information d, user information b, and environment information c output from the message content management unit 30. Then, based on the object type information d, user information b, and environment information c, it searches the object database 51A stored in the object storage unit 51 and obtains object information f as a search result from the object database 51A. Then, it outputs the obtained object information f to the message content management unit 30. This object information f is included in the message h that is ultimately output to the user, as will be described later.

[0036] Figure 3 is a data structure diagram showing an example of an object database.

[0037] The object database 51A stores object type information d, user information b, environment information c, and object information f, which indicates the content of an object determined by a combination of user information b, object type information d, and environment information c, in an associated manner. In the example shown in Figure 3, the target user, which is information corresponding to user information b, is stored as an example of user information b. Also, a URL is shown as an example of object information f, but object information f is not limited to URLs; for example, if it is a real-world object, it may be location information or more detailed object information other than a URL.

[0038] The second object acquisition unit 50 searches the object database 51A based on object type information d, user information b, and environment information c, acquires object information f as a search result, and outputs it to the message content management unit 30. This object information f is useful for promoting user actions.

[0039] Furthermore, the method by which the second object acquisition unit 50 acquires object information f is not limited to using the object database 51A, but may also be one that does not use the object database 51A.

[0040] For example, the second object acquisition unit 50 may, instead of the object database 51A, perform a search on a database containing the necessary resources, such as a publicly available web search engine, a store database, or a video service search system, in order to acquire object information f, and acquire object information f as a result. Note that information on the connection destinations, such as the web search engine, store database, or video service search system, can be stored in the object information database 41A along with object type information d.

[0041] The message content management unit 30 receives object information f output from the second object acquisition unit 50. It then outputs the object information f, along with the action content a, user information b, and environment information c, to the message generation unit 60.

[0042] The message generation unit 60 receives the action content a, user information b, environment information c, and object information f output from the message content management unit 30. Based on the action content a, user information b, environment information c, and object information f, and based on the text information g obtained from the large-scale language model 61A stored in the large-scale language model storage unit 61, it generates a message h that encourages the user to take action.

[0043] The large-scale language model 61A is a language model capable of generating text from a prompt. The large-scale language model 61A can utilize a model that has been trained to output text information g as a response to a prompt, for example, based on language models such as GTP (Non-Patent Literature 4) or T5 (Non-Patent Literature 5).

[0044] Figure 4 shows an example of a prompt when querying a large-scale language model.

[0045] The message generation unit 60 constructs a prompt, as illustrated in Figure 4, based on the action content a, user information b, environment information c, and object information f, in order to query the large-scale language model 61A. In response, the large-scale language model 61A outputs text information g as a response to the prompt.

[0046] The message generation unit 60 identifies the message portion contained in the output text information g and outputs the identified message portion as message h to the message return unit 70. This message h is a message intended to encourage the user to take action. Thus, the object information f is the data necessary for generating the message h that encourages the user to take action.

[0047] The message return unit 70 receives the message h output from the message generation unit 60 and sends it back to the client terminal 90. The message return unit 70 can also, for example, send the message h back to the client terminal 90 during the time period specified by the environment information c.

[0048] The client terminal 90 receives message h sent from the message return unit 70. The user can view message h using the client terminal 90 or via software and recognize the content of the awareness campaign to encourage action.

[0049] As explained above, the target behavior promotion message generation device 10 first determines the type of object necessary to take action based on the content of the action and the user's situation, such as the stage of behavioral change. Next, it identifies specific objects of the necessary type based on personal information such as age, gender, and place of residence, and generates a message h that includes these objects.

[0050] Such a target behavior promotion message generation device 10 can be implemented using a computer such as a PC.

[0051] Figure 5 is a simplified diagram showing a typical computer hardware configuration.

[0052] Computer 100 receives some kind of input from the outside, processes it, and outputs the result to the outside. Input is handled by input device 120, and output is handled by output device 140. CPU (Central Processing Unit) 110 controls the entire flow of data and processing, and performs calculations and other processing. In Figure 5, solid arrows represent the flow of data, and dashed arrows represent the flow of control.

[0053] For computer 100 to perform processing, it needs a program that describes the processing steps. The program contains a series of instructions and the data used by those instructions, all of which are temporarily stored in memory 130. CPU 110 retrieves the instructions from memory 130 one by one, interprets them, and operates according to those instructions.

[0054] Therefore, the target action promotion message generation device 10 is implemented by the computer 100, through the operation of the CPU 110 according to a program stored in the memory 130.

[0055] In the target action promotion message generation device 10, the message request receiving unit 20 corresponds to the input device 120, and the message return unit 70 corresponds to the output device 140. Furthermore, the message content management unit 30, the first object acquisition unit 40, the second object acquisition unit 50, and the message generation unit 60 are realized by the CPU 110 operating according to a program stored in the memory 130.

[0056] Furthermore, although not shown in Figure 5, the computer 100 can incorporate a storage device such as an SSD (Solid State Drive) or HDD (Hard Disk Drive). The object information storage unit 41, the object storage unit 51, and the large-scale language model storage unit 61 can be realized by such an internal storage device. In addition, the object information storage unit 41, the object storage unit 51, and the large-scale language model storage unit 61 are not limited to these internal storage devices, but can also be realized in an external storage medium such as a USB (Universal Serial Bus) memory, or in an area provided in a storage system such as a database server located in the cloud. Note that in Figure 1, the object information storage unit 41, the object storage unit 51, and the large-scale language model storage unit 61 are represented as separate storage units for clarity, but the object information storage unit 41, the object storage unit 51, and the large-scale language model storage unit 61 may be physically constructed within the same storage device.

[0057] Next, an example of the operation of the target behavior promotion message generation device according to this embodiment, configured as described above, will be explained.

[0058] Figure 6 is a flowchart showing an example of operation performed by the target behavior promotion message generation device according to this embodiment.

[0059] The message request receiving unit 20 communicates with the client terminal 90 and receives from the client terminal 90 the action content a that the user should be prompted to take, user information b about the user, and environment information c (S1). The action content a, user information b, and environment information c are the information necessary for the message generation unit 60 to generate message h.

[0060] The message request receiving unit 20 outputs the received action details a, user information b, and environment information c to the message content management unit 30 (S2).

[0061] The message content management unit 30 receives the action content a, user information b, and environment information c output by the message request receiving unit 20, and outputs them to the first object acquisition unit 40 (S3).

[0062] The first object acquisition unit 40 searches the object information database 41A stored in the object information storage unit 41 based on the action content a, user information b, and environment information c output from the message content management unit 30, and obtains object type information d as a search result from the object information database 41A, which indicates the type of object that should be provided to the user to facilitate the action, and outputs the obtained object type information d to the message content management unit 30 (S4).

[0063] The message content management unit 30 receives object type information d output from the first object acquisition unit 40 and outputs the received object type information d together with user information b and environment information c to the second object acquisition unit 50 (S5).

[0064] The second object acquisition unit 50 receives object type information d, user information b, and environment information c output from the message content management unit 30, searches the object database 51A stored in the object storage unit 51 based on the object type information d, user information b, and environment information c, obtains object information f as a search result from the object database 51A, and outputs it to the message content management unit 30 (S6).

[0065] The message content management unit 30 receives object information f output from the second object acquisition unit 50. Then, it outputs the object information f, along with the action content a, user information b, and environment information c, to the message generation unit 60 (S7).

[0066] The message generation unit 60 constructs a prompt, as illustrated in Figure 4, based on the action content a, user information b, environment information c, and object information f, in order to query the large-scale language model 61A. In response, the large-scale language model 61A outputs text information g that answers the prompt. The message generation unit 60 identifies the message portion contained in the output text information g and outputs the identified message portion as message h to the message return unit 70 (S8).

[0067] The message return unit 70 sends message h back to the client terminal 90 (S9).

[0068] The client terminal 90 receives message h sent from the message return unit 70. The user can view message h using the client terminal 90 or via software and recognize the content of the awareness campaign to encourage action.

[0069] As described above, the target behavior promotion message generation device 10 according to this embodiment can automatically generate messages that are highly effective in promoting behavior by identifying information that is likely to prompt specific actions according to the user's situation. This makes it possible, for example, to automatically generate messages that are highly effective in promoting specific health behaviors according to an individual's health status.

[0070] Furthermore, this message can be delivered to users through smartphone apps and other means. Because this message contains precise advice for better behavior and information encouraging immediate action, it can prompt users who receive the message to take effective action quickly.

[0071] In the above explanation, we described an example of operation based on action content a, user information b, and environment information c input from the client terminal 90. However, environment information c is optional, and even if environment information c is not input, the target action promotion message generation device 10 can generate message h by operating as described above using action content a and user information b, and send it back to the client terminal 90.

[0072] The present invention is not limited to the embodiments described above, and in the implementation stage, the components can be modified and implemented without departing from the gist of the invention. That is, the hardware configuration described in the embodiments above is merely an example, and for example, a GPU can be used for computation. Furthermore, various inventions can be formed by appropriately combining the multiple components disclosed in the embodiments above. For example, some components may be deleted from all the components shown in the embodiments. Furthermore, components from different embodiments may be appropriately combined. Also, environmental information c is optional. [Explanation of Symbols]

[0073] 10. Target Action Promotion Message Generation Device 20 Message Request Receiving Unit 30 Message Content Management Department 40. First Object Acquisition Unit 41 Object Information Storage Unit 41A Object Information Database 50 Second Object Acquisition Unit 51 Object Storage Unit 51A Object Database 60 Message generation unit 61 Large-scale language model memory 61A Large-scale language models 70 Message Return Department 90 client terminals 100 Computers 120 Input devices 130 memory 140 Output device a Action details b. User information c Environmental information d Object type information d1 Object type information (major category) d2 Object type information (subcategory) f Object Information g Text information h message

Claims

1. A first object acquisition unit acquires object type information from a pre-prepared object information database that indicates the type of object to be provided to the user in order to facilitate the action, based on the content of the action to be encouraged to the user and user information about the user. A second object acquisition unit acquires object information indicating the content of the object from a pre-prepared object database based on the user information and the object type information, A message generation unit generates a message to encourage the user to perform the action, using a pre-prepared language model based on the content of the action, the user information, and the object information. A target action promotion message generation device equipped with the following features.

2. The object information database stores the content of the action, the user information, and the object type information corresponding to the combination of the content of the action and the user information, in association with each other. A target behavior promotion message generation device according to claim 1.

3. The object database stores user information, object type information, and object information indicating the content of the object determined by a combination of the user information and the object type information, in an associated manner. A target behavior promotion message generation device according to claim 1.

4. The aforementioned actions are actions aimed at health. A target behavior promotion message generation device according to claim 1.

5. The aforementioned language model is trained to output text information in response to a specified prompt. The aforementioned message includes the text information, A target behavior promotion message generation device according to claim 1.

6. A method for generating goal action promotion messages, The processor, Based on the content of the action to be encouraged from the user and user information about the user, object type information indicating the type of object to be provided to the user in order to facilitate the said action is obtained from a pre-prepared object information database. Based on the user information and the object type information, object information indicating the content of the object is obtained from a pre-prepared object database. Based on the content of the aforementioned action, the user information, and the object information, a pre-prepared language model is used to generate a message that encourages the user to perform the aforementioned action. A method for generating messages that encourage target behavior.

7. A program for causing a computer to function as one of the components of the target behavior promotion message generation device according to any one of claims 1 to 5.

Images