Central server device, persona learning method, and program

Abstract

Provided is a central server device capable of learning a persona desired by a client even when the amount of data of a company owned by the client is small. A central server device according to the present invention includes: an information acquisition unit that receives, from a client device controlled by a client, at least one of client member information and a desired persona, the client member information concerning a member belonging to the client, and the desired persona representing a member image desired by the client, that determines requested member information specifications specifying member information to be requested, that acquires, from an external server group controlled by business operators serving as partners, partner member information concerning members belonging to the partners and matching the requested member information specifications, and that generates training data for learning a persona on the basis of at least the partner member information; and a persona learning unit that learns the persona on the basis of the training data.

Description

Central Server Device, Persona Learning Method, Program

[0001] The present disclosure relates to a central server device, a persona learning method, and a program for assisting marketing.

[0002] Conventionally, there is known a technique for generating a persona by inputting big data such as user attributes, related keywords, concerns, and interests into a generative AI (Non-Patent Document 1). A persona refers to a fictional portrait of a person who is the target of a product or service.

[0003] LY Corporation, "Quickly and Deeply Understand Personas by Integrating with Generative AI! Introduction to AI Interviews", [online], September 19, 2023, LY Corporation, [searched on October 17, 2024], Internet <URL: https: / / ds.yahoo.co.jp / report / 20230919.html>

[0004] Those in charge of product development and marketing require a huge amount of time and effort to obtain insights (questionnaire collection, group interviews). The person in charge wants to obtain insights based on the membership information the company has, but it is difficult to construct a persona if the amount of data the company has is small. Even if they try to use the membership information of other companies, there are high hurdles from the perspectives of personal information and security.

[0005] Therefore, an object of the present disclosure is to provide a central server device that can learn a persona desired by a client even when the amount of the client's own data is small.

[0006] The central server device of the present disclosure includes an information acquisition unit and a persona learning unit.

[0007] The information acquisition unit receives either client member information, which is information about members belonging to the client, or a desired persona, which is a persona representing the member profile desired by the client, from a client device controlled by the client. It then determines the requested member information specification, which is the specification for the member information to be requested. The unit acquires partner member information, which is information about members belonging to the partner that matches the requested member information specification, from a group of external servers controlled by the partner business operator. Based on this information, the unit generates training data for learning a persona. The persona learning unit learns a persona based on the training data.

[0008] According to the central server device described in this disclosure, even if the amount of data the client possesses is small, it is possible to learn the persona desired by the client.

[0009] A block diagram showing the device configuration of the marketing support system in Example 1. A flowchart showing the operation of the client device in Example 1. A flowchart showing the persona learning operation of the central server device in Example 1. A block diagram showing the functional configuration of the information acquisition unit of the central server device in Example 1. A flowchart showing the learning data generation operation of the central server device in Example 1. A diagram showing an example of clustering for generating requested member information specifications. A diagram showing an example of integrating client member information and partner member information. A diagram showing an example of generating an Instruction dataset based on integrated member information. A diagram showing an example of persona generation based on partner member information collected based on desired personas. A flowchart showing the persona selection operation of the central server device in Example 1. A diagram showing an example of a table for performing persona selection. A flowchart showing the response acquisition and transmission operation of the central server device in Example 1. A block diagram showing the functional configuration of the response acquisition unit of the central server device in Example 1. A flowchart showing the operation of the response acquisition unit of the central server device in Example 1. A diagram showing an example of the computer's functional configuration.

[0010] The embodiments of this disclosure will be described in detail below. Components having the same function will be numbered the same, and redundant explanations will be omitted.

[0011] The device configuration of the marketing support system 1000 of Embodiment 1 will be described below with reference to Figure 1. As shown in the figure, the marketing support system 1000 of this embodiment includes a client device 1, a central server device 2, and a group of external servers 3.

[0012] Client device 1 is a device controlled by the client and includes a member information storage unit 10 that stores client member information, which is information about members belonging to the client.

[0013] The central server device 2 includes an information acquisition unit 21, a persona learning unit 22, a model / persona storage unit 23, and an answer acquisition unit 24.

[0014] The external server group 3 includes business server devices 3-1, ..., 3-N controlled by business partners 1, ..., N (where N is an integer of 2 or more), and each business server device 3-1, ..., 3-N includes a member information storage unit 31-1, ..., N that stores partner member information, which is information about members belonging to the partner.

[0015] The following sections will provide further explanations regarding the client, central server device 2, and partner.

[0016] <Client> The client has the following characteristics, for example:

[0017] The client is a user of Marketing Support System 1000 and conducts development and marketing operations related to its own products (services). The client uses Marketing Support System 1000 to perform tasks such as surveys, in-depth interviews, and group discussions in order to gain deep insights into customers of its products (services).

[0018] When a client uses the marketing support system 1000, they transmit all or part of their member information (client member information) or a desired persona to the central server device 2.

[0019] When using the marketing support system 1000, the client specifies to the central server device 2 the task they want to perform (e.g., survey, in-depth interview, group discussion), the number of attempts, and the model (e.g., LLM (Large Language Models), tsuzumi, ChatGPT, etc.).

[0020] The client obtains the results of their desired tasks from the marketing support system 1000 and uses them in the development and marketing of their own products (services).

[0021] The client pays a fee for using the marketing support system 1000.

[0022] <Central Server Device 2> Central Server Device 2 has the following features, for example.

[0023] The central server device 2 provides information that contributes to product development and marketing based on input from clients.

[0024] The central server device 2 clusters the client member information sent by the client and creates a requested member information specification (classification result of client member information).

[0025] The central server device 2 collects the necessary partner member information based on the requested member information specifications, and integrates and consolidates the member information of each company.

[0026] The central server device 2 generates personas by supervising a specified model (e.g., LLM) using supervised learning (e.g., Instruction-Tuning) based on the integrated and matched member information.

[0027] The central server device 2 generates the necessary prompts based on the tasks entered by the client.

[0028] The central server device 2 inputs prompts into the generated persona and outputs the marketing information requested by the client.

[0029] <Partner> A partner might have the following characteristics, for example:

[0030] • Partners are businesses that provide partner member information necessary for persona generation, and are affiliated with the marketing support system 1000.

[0031] The partner will provide the necessary partner member information based on a request from the central server device 2.

[0032] Partners receive compensation based on the partner member information they provide.

[0033] The types of businesses that could potentially become partners are diverse. For example, telecommunications companies, e-commerce companies, media companies, and event organizers are some possibilities.

[0034] The operation of client device 1 will be explained below with reference to Figure 2. First, client device 1 accesses central server device 2 (S11). Next, client device 1 transmits at least one of the following: client member information or a desired persona, which is a persona representing the member profile desired by the client (S12). Next, client device 1 specifies the requested task to central server device 2 (S13).

[0035] Furthermore, client device 1 specifies to central server device 2 the number of attempts (let's call it n, where n is a natural number) for the requested task (S14). Furthermore, client device 1 specifies to central server device 2 the model to be used for persona generation (S15). Steps S14 and S15 are optional operations and may be omitted. If steps S14 and S15 are omitted, the specification of the number of task attempts and the model can be performed by, for example, central server device 2.

[0036] Client device 1 receives the answer to the task from central server device 2 (S16). Details of persona learning and answer acquisition operations will be described later.

[0037] The persona learning operation of the central server device 2 will be explained below with reference to Figure 3. First, the information acquisition unit 21 receives at least one of either client member information or a desired persona from the client device 2, determines the requested member information specification which is the specification of the member information to be requested, acquires partner member information that matches the requested member information specification from the external server group 3, and generates learning data for learning a persona based at least on the partner member information (S21). The persona learning unit 22 learns a persona based on the learning data (S22).

[0038] The following describes a detailed example of the functional configuration of the information acquisition unit 21 with reference to Figure 4. As shown in the figure, the information acquisition unit 21 includes a requirements specification determination unit 211, a partner member information acquisition unit 212, a learning data generation unit 213, and a persona selection unit 214. Note that the persona selection unit 214 is optional and can be omitted in some cases, so it is shown with a dashed line in the figure.

[0039] The learning data generation operation of the information acquisition unit 21 will be further explained below with reference to Figure 5, based on the functional configuration example in Figure 4. The request specification determination unit 211 receives at least one of either client member information or a desired persona from the client device 2 and determines the request member information specification, which is the specification of the member information to be requested (S211A).

[0040] The requested member information specification is the classification result of the client member information, or the desired persona itself.

[0041] As mentioned above, client member information refers to information about members belonging to the client. The client member information received from client device 2 may be all of the client member information owned by the client, or it may be only a part of it.

[0042] The client member information received in step S211A is information used to determine the requested member information specifications. Therefore, when the central server device 2 receives all of the client member information owned by the client, it means that the client aims to gain some knowledge about all of the member information it possesses.

[0043] Also, when the central server device 2 receives a part of the client member information owned by the client, it means that the client aims to obtain some knowledge about a part of the member information it has.

[0044] Also, as described above, the desired persona is a persona representing the member image desired by the client, and is represented by setting ranges for attributes such as age and gender. For example, it can be expressed as desired persona = [living in Tokyo, annual income of X yen or more, in their 30s, male], etc. When the desired persona is received from the client, it means that the client aims to obtain some knowledge about the member image having the attributes indicated by the desired persona.

[0045] When the client member information is received from the client device 1, the required member information specification is generated as shown in FIG. 6, for example. In the example of FIG. 6, two processing options are illustrated.

[0046] In processing option 1, the client member information is clustered by the k-means method (k = 6) by age and gender, and clustering results such as... names of males aged 20 - 24, ... names of males aged 25 - 29, ... names of males aged 30 - 44, ... names of males aged 45 - 55, ... names of females aged 18 - 28, ... names of females aged 31 - 35 are obtained. This clustering result is the required member information specification.

[0047] In processing option 2, classification is performed based on marketing attributes (C, T, M1 - M4, F1 - F4, for example, the C layer is males and females aged 4 - 12, the F1 layer is females aged 20 - 34), and classification results such as... names of C, ... names of T, ... names of M1, ... names of M2, ... are obtained. This classification result is the required member information specification.

[0048] Next, the partner member information acquisition unit 212 acquires partner member information that matches the requested member information specification from the external server group 3 (S212). Here, "matches" means that the ratio for each attribute indicated by the classification result of the client member information is approximately the same as the ratio for each attribute of the acquired partner member information. Or "matches" indicates that the range of attributes indicated by the desired persona matches the range of attributes of the acquired partner member information.

[0049] Next, the learning data generation unit 213 generates learning data for learning a persona based at least on the partner member information (S213).

[0050] For example, when client member information is received from the client device 1, learning data is generated based on the integrated data of the client member information and the partner member information. On the other hand, when only the desired persona is received from the client device 1, there may be no client member information. In this case, since learning data is generated based only on the partner member information, step S213 indicates that learning data is generated "at least" based on the partner member information.

[0051] FIG. 7 is a diagram showing an example in which the learning data generation unit 213 integrates client member information and partner member information as learning data to generate learning data. An example of integrating client member information and the partner member information of Business Operator 1 and Business Operator 2 is shown in the figure.

[0052] When integrating the member information, the learning data generation unit 213 shall handle the labels of each member information (referring to the item names in the database, such as ID, age,... etc.) such that the same labels are unified in the same column and new columns are added for different labels. Also, it is preferable for the learning data generation unit 213 to group the member information based on information such as the phone number so as not to duplicate the information.

[0053] In this embodiment, Instruction tuning is used for persona learning. The learning data generation unit 213 converts the integrated member information into a dataset (Instruction dataset) necessary for learning with Instruction tuning. For example, the following method can be used to generate an Instruction dataset for one cluster.

[0054] - Create a template sentence in Instruction that asks for the most frequent label among those not used in the cluster, based on the content of the item, and insert the parameter. - Input should be data with duplicates removed for each label. - Output should be calculated using arithmetic operations and inserted into the template sentence. Figure 8 shows an example of how the integrated member information in Figure 7 was converted into the Instruction dataset based on this method.

[0055] Next, Figure 9 shows an example of training data generation when the requested member information specification = desired persona. In the example shown in the figure, it is assumed that the requested member information specification = desired persona = [30s, male] is specified. In this case, the partner member information acquisition unit 212 selects and collects member information that has attributes matching the desired persona from the partner member information of business operator 1, business operator 2, business operator 3, ... The training data generation unit 213 integrates the collected partner member information.

[0056] In this case, the learning data generation unit 213 preferably unifies the same labels into the same column, adds a new column for different labels, and then performs data matching, similar to the example in Figure 7. As mentioned above, the learning data generation unit 213 converts the integrated member information into text and then into an Instruction dataset, which is then sent to the persona learning unit 22 as learning data.

[0057] The persona learning unit 22 acquires training data and learns personas by supervising the model (LLM, tsuzumi, ChatGPT, etc.) through supervised learning (instruction tuning in this embodiment) (S22). In the example shown in the figure, the model is supervised learning based on the desired persona = [30s, male], and a persona of a 30-year-old male is generated.

[0058] The personas that are learned refer to models (such as LLMs) that reflect the tendencies of attributes shown in the learning data (e.g., age, gender, place of residence, services used, purchasing trends, hobbies and preferences, etc.).

[0059] Next, with reference to Figures 10 and 11, the optional operation of persona selection will be explained. In this optional operation, persona learning is omitted and previously learned personas are reused. Specifically, multiple types of previously learned personas are pre-stored in the model / persona memory unit 23. The persona selection unit 214 determines the requested persona specification, which is the specification of the persona to be requested, based on at least one of the client member information or the desired persona (S211B), and selects a persona from the model / persona memory unit 23 that matches the requested persona specification (S214).

[0060] Figure 11 illustrates the required persona specifications and the personas that match those specifications. As shown in the figure, if the required persona specifications fall under category 1, previously trained persona A that matches type 1 is selected. Similarly, if the required persona specifications fall under categories 2, 3, ..., previously trained personas B, C, ... that match types 2, 3, ... are selected.

[0061] Next, as shown in Figure 12, when the response acquisition unit 24 acquires a task from the client device 1, it acquires a response to the task based on the learned (or selected) persona and transmits the response to the client device 1 (S24). The timing of task acquisition may be at the same time as the client member information / desired persona, or at any other time after the client member information / desired persona has been received.

[0062] Furthermore, some tasks are performed based on multiple personas (for example, group interviews). In this case, the persona learning unit 22 learns the multiple personas necessary for the task, and the response acquisition unit 24 acquires responses from the multiple personas as answers.

[0063] For example, as shown in Figure 13, the response acquisition unit 24 may be configured to include a prompt generation unit 241, a response storage unit 242, a statistical processing unit 243, and an advice generation unit 244.

[0064] When the client device 1 receives a task requiring multiple personas and specifies repeated execution of the task over a predetermined number of attempts (n times), the prompt generation unit 241 first generates prompts corresponding to the task (prompts 1, 2, 3, ..., with the numbers corresponding to each persona) (S241).

[0065] The response storage unit 242 repeatedly performs the operation (S242) of acquiring and storing the response of each persona to each prompt for a predetermined number of attempts. The number in parentheses represents the number of attempts, and the number outside the parentheses represents the number of the corresponding persona. Responses 1(1), Response 2(1), Response 3(1), ..., Response 1(n), Response 2(n), Response 3(n), ... are acquired and stored. The stored responses are output as a list of responses.

[0066] The statistical processing unit 243 performs statistical processing based on the responses stored in the response storage unit 242 and outputs the statistical results (S243). An example of the statistical results being expressed in question and answer text format is shown below.

[0067] Example question: Are you interested in this product? What do you think is the best medium for advertising? Example answer: X percent of the F1 demographic responded that they were interested. The media that the F1 demographic interacts with most frequently are YouTube, TV, etc., in that order, making YouTube advertising the best option. ... The advice generation unit 244 generates advice based on the statistical results and outputs the advice (S244). An example of the advice converted into text format is shown below.

[0068] Example advice: The most effective target audience for this product is women in their 20s (F1 demographic). The current marketing strategies and advertising media do not match the target audience, so utilizing YouTube and TV commercials would be more effective. ... According to the marketing support system 1000 and central server device 2 of this embodiment, clients and partners can generate personas that reflect each other's member information and execute tasks without directly exchanging member information with each other, thus providing high convenience for clients.

[0069] In this case, the partner and client do not need to exchange member information directly, thus ensuring a high level of confidentiality of member information.

[0070] Furthermore, the marketing support system 1000 and central server device 2 of this embodiment can streamline product development and marketing operations, and provide customer insights that cannot be obtained through normal activities. It also lowers the barriers to data utilization and promotes data distribution. By performing the acquisition of member information and its use as learning data in a secure computing area, risks such as information leaks can be reduced. Additionally, partners can generate revenue by utilizing member information.

[0071] <Note> The functions realized by the components described herein may be implemented in a circuitry or processing circuitry, including general-purpose processors, application-specific processors, integrated circuits, ASICs (Application Specific Integrated Circuits), CPUs (a Central Processing Unit), conventional circuits, and / or combinations thereof, programmed to realize the functions described herein. A processor includes transistors and other circuits and is considered a circuitry or processing circuitry. A processor may be a programmed processor that executes a program stored in memory.

[0072] In this specification, circuitry, unit, and means are hardware programmed to perform or execute the functions described herein. Such hardware may be any hardware disclosed herein, or any hardware known to be programmed to perform or execute the functions described herein.

[0073] If the hardware is a processor that is considered to be a type of circuitry, then the circuitry, means, or unit is a combination of hardware and software used to constitute the hardware and / or processor.

[0074] The various processes described above can be carried out by loading a program that executes each step of the above method into the recording unit 10020 of the computer 10000 shown in Figure 14, and then causing the control unit 10010, input unit 10030, output unit 10040, etc. to operate.

[0075] The program describing this process can be recorded on a computer-readable recording medium. Any computer-readable recording medium can be used, such as a magnetic recording device, optical disc, magneto-optical recording medium, or semiconductor memory.

[0076] Furthermore, this program may be distributed, for example, by selling, transferring, or lending portable recording media such as DVDs or CD-ROMs on which the program is recorded. Alternatively, the program may be stored in the storage device of a server computer and distributed by transferring the program from the server computer to other computers via a network.

[0077] A computer executing such a program may, for example, first store the program recorded on a portable storage medium or a program transferred from a server computer in its own storage device. Then, when processing is to be executed, the computer reads the program stored on its own storage medium and executes the processing according to the read program. Alternatively, the computer may directly read the program from the portable storage medium and execute the processing according to that program, or it may sequentially execute the processing according to the received program each time a program is transferred to it from a server computer. Furthermore, the processing may be executed using a so-called ASP (Application Service Provider) type service, where the processing function is realized only by issuing execution instructions and obtaining results, without transferring the program from the server computer to this computer.In addition, the processing may be executed using a so-called SaaS (Software as a Service) type service, where a part of the server computer is made available to the user along with the program. Furthermore, the term "program" in this form includes information used for processing by an electronic computer that is equivalent to a program (data, etc., that is not a direct instruction to the computer but has the property of defining the processing of the computer).

[0078] The following additional information is disclosed regarding the embodiments described above.

[0079] (Note 1) A central server device that includes a memory and at least one processor connected to the memory, wherein the processor receives at least one of either client member information, which is information about members belonging to the client, or a desired persona, which is a persona representing a member profile desired by the client, from a client device controlled by the client, determines a requested member information specification, which is the specification of the member information to be requested, acquires partner member information, which is information about members belonging to the partner that matches the requested member information specification, from a group of external servers controlled by a partner business operator, generates learning data for learning the persona based at least on the partner member information, and learns the persona based on the learning data.

[0080] (Note 2) A non-temporary storage medium storing a program executable by a computer to perform a persona learning process, wherein the persona learning process receives at least one of either client member information, which is information about members belonging to the client, or a desired persona, which is a persona representing a member profile desired by the client, from a client device controlled by the client, determines a requested member information specification, which is the specification of the member information to be requested, obtains partner member information, which is information about members belonging to the partner that matches the requested member information specification, from a group of external servers controlled by a partner business operator, generates learning data for learning the persona based at least on the partner member information, and learns the persona based on the learning data.

[0081] (Note 3) A central server device as described in Note 1, wherein the processor stores multiple types of personas learned in the past, determines a requested persona specification which is the specification of a persona to be requested based on at least one of the client member information or the desired persona, and selects a persona that matches the requested persona specification.

[0082] (Appendix 4) A non-temporary storage medium as described in Appendix 2, wherein the persona learning process stores multiple types of personas learned in the past, determines a requested persona specification which is the specification of a persona to be requested based on at least one of the client member information or the desired persona, and selects a persona that matches the requested persona specification.

[0083] (Appendix 5) A central server device as described in Appendix 1, wherein the processor, when it obtains a task from the client device, obtains an answer to the task based on the learned persona and transmits the answer to the client device.

[0084] (Appendix 6) A non-temporary storage medium according to Appendix 2, further storing a program executable by a computer to perform a response process to a task, wherein the response process, when the task is obtained from the client device, obtains a response to the task based on the learned persona and transmits the response to the client device.

[0085] (Appendix 7) A central server device as described in Appendix 5, wherein the processor learns the multiple personas necessary for the task when it receives a task to be executed based on a plurality of personas from the client device, and when the client device specifies the execution of the task over a predetermined number of trials, it repeatedly obtains responses from the plurality of personas over a predetermined number of trials, obtains a list of the responses, statistical results based on the responses, and advice based on the statistical results as the answer, and transmits them to the client device.

[0086] (Appendix 8) A non-temporary storage medium as described in Appendix 6, wherein the response processing learns the multiple personas necessary for the task when it receives the task to be performed based on a plurality of personas from the client device, and when the client device specifies the execution of the task over a predetermined number of trials, it repeatedly obtains responses from the plurality of personas over a predetermined number of trials, and obtains a list of the responses, statistical results based on the responses, and advice based on the statistical results as the response and transmits them to the client device.

[0087] 1 Client device 10 Member information storage unit 2 Central server device 21 Information acquisition unit 22 Persona learning unit 23 Model / persona storage unit 24 Response acquisition unit 3 External server group 3-1, ..., N Business operator server device 31-1, ..., N Member information storage unit 211 Requirements specification determination unit 212 Partner member information acquisition unit 213 Learning data generation unit 214 Persona selection unit 241 Prompt generation unit 242 Response storage unit 243 Statistical processing unit 244 Advice generation unit 10000 Computer 10010 Control unit 10020 Recording unit 10030 Input unit 10040 Output unit 10050 Display unit

Claims

1. A central server device including an information acquisition unit that receives at least one of the following from a client device controlled by the client: client member information, which is information about members belonging to the client, or a desired persona, which is a persona representing the member profile desired by the client; a persona learning unit that learns the persona based on the learning data.

2. A central server device according to claim 1, comprising a model / persona storage unit that stores a plurality of types of personas learned in the past, wherein the information acquisition unit determines a requested persona specification, which is the specification of a persona to be requested, based on at least one of the client member information or the desired persona, and selects a persona from the model / persona storage unit that matches the requested persona specification.

3. A central server device according to claim 1, comprising an answer acquisition unit that, when it acquires a task from a client device, acquires an answer to the task based on the learned persona and transmits the answer to the client device.

4. A central server device according to claim 3, wherein the persona learning unit learns the multiple personas necessary for a task when it receives a task to be performed based on a plurality of personas from the client device, and the response acquisition unit repeatedly acquires responses from the plurality of personas over a predetermined number of trials when the client device specifies the execution of the task over a predetermined number of trials, and acquires a list of the responses, statistical results based on the responses, and advice based on the statistical results as the response and transmits them to the client device.

5. A persona learning method executed by a central server device, comprising the steps of: receiving at least one of either client member information, which is information about members belonging to the client, or a desired persona, which is a persona representing a member profile desired by the client, from a client device controlled by the client, determining a requested member information specification, which is the specification of the member information to be requested; obtaining partner member information, which is information about members belonging to the partner that matches the requested member information specification, from a group of external servers controlled by a partner business operator, and generating learning data for learning the persona based at least on the partner member information; and learning the persona based on the learning data.

6. A program that causes a computer to function as the central server device described in claim 1.

Images