Generating device, generating method, and program for simulating answers to questions

Abstract

A generating device, a generating method, and a program are provided that are capable of obtaining a variety of answers for evaluating the validity and appropriateness of a question in advance. [Solution] The generation device 100 includes a classification unit 120 that classifies candidates who can answer questions into multiple clusters based on characteristic information about the candidates, a calculation unit 130 that determines representative characteristics that indicate the characteristics of a representative who represents each cluster based on the characteristic information about the candidates included in each classified cluster, a reception unit 140 that receives a question text, and a generation unit 150 that causes a generation AI server 300 that simulates a representative having the representative characteristics determined for each cluster to generate a simulated answer to the question text received by the reception unit 140.

Description

[Technical Field]

[0001] The present invention relates to a generating device, a generating method, and a program for simulating and generating answers to questions. [Background technology]

[0002] Surveys such as interviews and questionnaires are conducted in various fields targeting a large number of users via the Internet. As an example, Patent Document 1 discloses a system that can realize a digital clone survey in order to conduct a survey faster and at a lower cost than conventional surveys.

[0003] This system uses personalized AI that learns an individual's expressions and generates sentences based on those expressions to output answers to survey questions. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2022-32935 Summary of the Invention [Problem to be solved by the invention]

[0005] Incidentally, after conducting a survey like the one above, it is sometimes only after looking at the collected responses that one realizes areas for improvement, such as whether the question should have been phrased differently or whether certain questions should have been included together. In such cases, it is necessary to revise the questions and conduct the survey again. For this reason, there was a desire to obtain a diverse sample of responses in order to evaluate the effectiveness and appropriateness of the questions before conducting the survey.

[0006] The present invention has been made in consideration of the above-mentioned situation, and aims to provide a generation device, a generation method, and a program that can obtain a variety of answers to evaluate the effectiveness and appropriateness of questions in advance. [Means for solving the problem]

[0007] In order to solve the above problems, a generating device according to the present invention comprises: a classification unit that classifies candidates who can answer questions into a plurality of clusters based on characteristic information of the candidates; a calculation unit that calculates a representative characteristic that indicates a characteristic of a representative that represents each of the clusters based on the characteristic information of the candidates included in each of the classified clusters; a reception unit for receiving questions; a generation unit that causes a generation AI that simulates a representative having the representative characteristics obtained for each of the clusters to generate a simulated answer to the question sentence received by the reception unit; Equipped with. [Effects of the Invention]

[0008] According to the present invention, it is possible to provide a generation device, a generation method, and a program that are capable of obtaining a variety of answers for evaluating the validity and appropriateness of a question in advance. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 10 is an explanatory diagram showing cooperation between the generating device and other devices. [Figure 2] FIG. 2 is an explanatory diagram illustrating the functional configuration of a generating device. [Figure 3] 3 is a diagram showing an example of an attribute table stored in a user DB shown in FIG. 2. FIG. [Figure 4] 3 is a diagram showing an example of a behavior history table stored in a user DB shown in FIG. 2. FIG. [Figure 5] 3 is a diagram showing an example of a classification result table generated by the classification unit shown in FIG. 2. FIG. [Figure 6] 3 is a diagram showing an example of a representative characteristic table generated by a calculation unit shown in FIG. 2. FIG. [Figure 7] FIG. 2 is an explanatory diagram illustrating the physical configuration of the generation device. [Figure 8]10 is a flowchart of a candidate classification process performed by the generation device. [Figure 9] 10 is a flowchart of an answer generation process performed by the generation device. [Figure 10] 3 is a diagram showing an example of a prompt generated by the generating unit shown in FIG. 2. FIG. [Figure 11] FIG. 10 is a diagram illustrating an example of a response table generated by a generating unit. DETAILED DESCRIPTION OF THE INVENTION

[0010] A generating device, a generating method, and a program according to an embodiment of the present invention will be described in detail with reference to the drawings. In the drawings, identical or corresponding parts are designated by the same reference numerals. This embodiment is for illustrative purposes only and does not limit the scope of the present invention. Therefore, those skilled in the art may employ embodiments in which each or all of these elements are replaced with equivalents, and these embodiments are also within the scope of the present invention.

[0011] (Overall composition) 1 is an explanatory diagram showing cooperation between a generating device 100 according to an embodiment of the present invention and other devices. As shown in the figure, the generating device 100 is communicably connected to a terminal 200 and a generation AI (Artificial Intelligence) server 300 via a communication network 400. Note that while one terminal 200 is shown in FIG. 1, the number of applicable terminals 200 is not limited to this, and multiple terminals 200 may be applied.

[0012] The generating device 100 is configured with one or more server computers. The generating device 100 is operated by a business operator that provides a crowdsourcing service that requests tasks, including responses to questionnaires, from an unspecified number of users.

[0013] The generation device 100 is a device that receives questionnaire questions and generates simulated answers to the received questions. Specifically, the generation device 100 classifies candidates who can answer the questionnaire into multiple clusters based on characteristic information of the candidates. The characteristic information includes, for example, attributes such as the candidate's age, gender, and occupation, and behavioral history such as product purchases and responses to past questionnaires.

[0014] The generation device 100 obtains, for each classified cluster, a representative characteristic that indicates the characteristics of a representative who represents the cluster. Based on the obtained representative characteristic for each cluster, the generation device 100 generates a prompt for each cluster that instructs the generation AI server 300 to simulate the representative and generate an answer to the question. The generation device 100 transmits each generated prompt to the generation AI server 300, causing it to simulate the generation of an answer to the question.

[0015] Terminal 200 is an information terminal (so-called computer) such as a PC (Personal Computer), tablet, or smartphone, and is used by, for example, a requester, such as a company or individual requesting the implementation of a survey through a crowdsourcing service. The requester uses terminal 200 to create survey questions, view mock answers generated by generation device 100, and perform tasks such as revising the questions.

[0016] The generation AI server 300 includes an AI that generates an answer based on a prompt that instructs the generation of an answer to a question. The generation AI server 300 includes, for example, a sentence generation AI that generates sentences, such as ChatGPT, GEMINI, Catchy, Notion AI, or other sentence generation AIs, or programs, services, or software that use these.

[0017] In addition, sentence generation AI models include any language model or large-scale language model, such as GPT, PaLM, LaMDA, LLaMa, Claude, OpenCALM, or language models or large-scale language models that have been modified, improved, transferred, or additionally learned from these.

[0018] The generation AI server 300 may further include an image generation AI and have the function of generating images.

[0019] Communications network 400 may include various types of networks, such as a local area network (LAN), a wide area network (WAN) such as the Internet, a telecommunications network such as the public switched telephone network (PSTN), a wireless network, a public switched network, a satellite network, a cellular network, a public land mobile network (PLMN), a metropolitan area network (MAN), a private network, an ad-hoc network, an intranet, an optical fiber-based network, or any combination of these or other types of networks.

[0020] (Functional configuration of the generating device) 2 is an explanatory diagram showing the functional configuration of the generating device 100. The generating device 100 includes a user DB (Database) 110, a classifying unit 120, a calculating unit 130, a receiving unit 140, and a generating unit 150.

[0021] The user DB 110 is a database that stores characteristic information of candidates who can answer questionnaires, and stores the candidates' responses to past questionnaires, the classification results of the candidates by the classification unit 120, and the representative characteristics calculated by the calculation unit 130. Candidates who can answer questionnaires are, for example, users registered with the crowdsourcing service. These users receive tasks provided by requesters online and perform the tasks. Note that, among all users registered with the crowdsourcing service, users who have not logged in or performed tasks for a certain period of time may be excluded from the candidates, or users who meet certain conditions may be extracted from all users registered with the crowdsourcing service and used as candidates.

[0022] Specifically, the user DB 110 includes an attribute table that stores attribute information of each candidate and a behavior history table that stores the behavior history of each candidate. Examples of the attribute table and the behavior history table are shown in FIGS. 3 and 4, respectively.

[0023] 3, the attribute table includes information such as a "user ID" that uniquely identifies a candidate, and "attributes" that indicate the attributes of each candidate, including their age, gender, occupation, family structure, etc. The attributes are not limited to the example shown in the figure, and may further include, for example, information such as the candidate's address, educational background, income, etc., or may be any other combination of information.

[0024] As shown in FIG. 4, the behavior history table includes the following information: a "user ID" that uniquely identifies a candidate; an "action type" that indicates the type of behavior performed by the candidate; a "product ID" that uniquely identifies a product purchased by the candidate; a "product category" that indicates the category of product purchased by the candidate; a "survey ID" that uniquely identifies a survey previously answered by the candidate; a "response ID" that includes information uniquely identifying the candidate's response; and a "timestamp" that indicates the date and time the behavior was performed. In the illustrated example, the "action type" includes either "purchase" or "survey response," but is not limited to these and may include any other behavior, such as viewing a product page, searching for a product, or registering a brand or shop as a favorite. Information regarding product purchases, etc., may be obtained from a management server (not shown) that manages the e-commerce service.

[0025] Returning to FIG. 2, the classification unit 120 classifies candidates who can answer the questionnaire into multiple clusters. Specifically, the classification unit 120 classifies multiple candidates stored in the user DB 110 into multiple clusters based on the characteristic information of each candidate. For example, the classification unit 120 compares feature vectors obtained by vectorizing the characteristic information (attributes and behavioral history) of each candidate with each other, and classifies based on the similarity and distance of the feature vectors. The classification method may be hierarchical or non-hierarchical. Any calculation method such as Ward's method, group average method, shortest distance method, or longest distance method can be used for hierarchical classification. Any calculation method such as k-means method can be used for non-hierarchical classification.

[0026] The classification unit 120 generates a classification result table indicating the classification results and stores the table in the user DB 110. An example of the classification result table is shown in Fig. 5. As shown in the figure, the classification result table is a table that associates a "user ID," which is information that uniquely identifies a candidate, with a "cluster ID," which is information that uniquely identifies a cluster classified by the classification unit 120.

[0027] Returning to FIG. 2, the calculation unit 130 determines representative characteristics that indicate the characteristics of a representative that represents each of the multiple clusters classified by the classification unit 120. For example, the calculation unit 130 generates a representative vector based on the feature vectors of the multiple candidates included in each cluster. The calculation unit 130 calculates, for example, the center of gravity of the feature vectors of all the candidates included in the cluster or the feature vector of the candidate closest to that center of gravity as the representative vector of each cluster. Based on the calculated representative vector, the calculation unit 130 generates a representative characteristic table that stores the representative characteristics for each cluster.

[0028] An example of the representative characteristic table is shown in Fig. 6. As shown in the figure, the representative characteristic table includes a "cluster ID" which is information that uniquely identifies a cluster, "attributes" which include the "age group," "gender," "occupation," and "family composition" of the representative of each cluster, and a "behavioral history" which includes the "product category" of the purchased product, the "purchase frequency" of the product, and "survey responses" which are information that uniquely identifies the representative's response data to surveys conducted in the past.

[0029] 2, the receiving unit 140 receives a question. Specifically, the receiving unit 140 waits for a question and receives the question sent from the terminal 200.

[0030] The generation unit 150 generates an answer to the question received by the reception unit 140. Specifically, the generation unit 150 generates a prompt for each cluster that instructs the generation AI server 300 to generate an answer to the question by simulating the representative, based on the representative characteristics of each cluster calculated by the calculation unit 130 and the received question.

[0031] The generation unit 150 sends the generated prompt to the generation AI server 300 and obtains the answer generated by the generation AI server 300. Details of the processing by the generation unit 150 will be described later. Note that the generation AI server 300 may be configured to be implemented as part of the function of the generation unit 150.

[0032] (Hardware configuration of information processing device) 7 is a block diagram showing the hardware configuration of the generating device 100. The generating device 100 includes a CPU 11 that executes processing according to a program, a RAM 12 that is a volatile memory, a ROM 13 that is a non-volatile memory, a storage unit 14 that stores data, an input unit 15 that accepts input of information, a display unit 16 that visualizes and displays information, and a communication unit 17 that transmits and receives information, all of which are connected via an internal bus 99.

[0033] The CPU 11 controls the overall operation of the generating device 100, is connected to each component, and exchanges control signals and data with each other. The CPU 11 executes various processes by reading programs stored in the storage unit 14 into the RAM 12 and executing them. The CPU 11 executes each process by the classification unit 120, calculation unit 130, reception unit 140, and generation unit 150 as the main functions provided by the programs.

[0034] The RAM 12 is used to temporarily store data and programs, and stores programs and data read from the storage unit 14, as well as other data necessary for communication. The RAM 12 is used as a work area for the CPU 11.

[0035] The ROM 13 stores a control program executed by the CPU 11 for the basic operation of the generating device 100, a BIOS (Basic Input Output System), and the like.

[0036] The storage unit 14 includes a hard disk drive, a flash memory, etc., and stores programs executed by the CPU 11 and various data used when the programs are executed. The storage unit 14 functions as a user DB 110.

[0037] The input unit 15 is a user interface including a touch panel, a keyboard, a mouse, a communication device, etc. The input unit 15 receives an operation input from a user of the generation device 100, and outputs a signal corresponding to the received operation input to the CPU 11.

[0038] The display unit 16 is a display device such as a liquid crystal display or an organic EL (Electro Luminescence) display that visualizes and displays information.

[0039] The communication unit 17 is a network termination device or a wireless communication device that connects to the network, and a serial interface or a LAN (Local Area Network) interface that connects to them. The generation device 100 communicates with the terminal 200, the generation AI server 300, etc. via the communication unit 17. The communication unit 17 functions as the reception unit 140.

[0040] (Candidate Classification Processing) Next, the operation of the generation device 100 will be described with reference to the drawings. First, the candidate classification process for classifying candidates who can answer a questionnaire into a plurality of clusters and calculating the representative characteristics of the representatives who represent each cluster will be described with reference to FIG.

[0041] The candidate classification process is started, for example, based on an execution instruction from an administrator of the generation device 100. Note that the generation device 100 may be configured to start the candidate classification process at a preset timing, such as daily, weekly, or monthly.

[0042] The classification unit 120 waits for an execution instruction, and if an execution instruction is received (step S101; Yes), the process proceeds to step S102. On the other hand, if an execution instruction is not received (step S101; No), the classification unit 120 waits for an execution instruction.

[0043] In step S102, the classification unit 120 acquires characteristic information of candidates who can answer the questionnaire (step S102). Specifically, the classification unit 120 accesses the user DB 110 and reads out the attribute table illustrated in FIG. 3 and the behavior history table illustrated in FIG. 4.

[0044] Next, the classification unit 120 classifies the candidates into multiple clusters (step S103). Specifically, the classification unit 120 classifies the candidates into multiple clusters based on the attributes and behavioral history of each candidate. For example, the classification unit 120 uses any classification method to calculate the similarity and distance between candidates based on feature vectors obtained by vectorizing the attributes and behavioral history of each candidate, and then classifies the candidates. Note that in the case of data that does not have numerical magnitude or order, such as categorical data such as gender or occupation or questionnaire responses, these can be converted into numerical values ​​to generate feature vectors. The classification unit 120 generates a classification result table showing the classification results, as shown in FIG. 5, and stores it in the user DB 110.

[0045] Next, the calculation unit 130 calculates the representative characteristics of the representatives representing each of the multiple clusters generated in step S103 (step S104). For example, the calculation unit 130 generates a representative vector based on the feature vectors of the candidates included in each cluster. For example, the calculation unit 130 calculates the center of gravity of the feature vectors of all candidates included in the cluster, or the feature vector of the candidate closest to the center of gravity, as the representative vector of each cluster. The calculation unit 130 calculates the center of gravity of the feature vectors by averaging the feature vectors of all candidates included in the cluster. Note that when calculating the representative characteristics of responses to a free-response questionnaire, feature vectors may be generated using any method, such as BoW (Bag of Words), TF-IDF, or word embedding, and the most frequently occurring words or phrases may be identified or representative topics for each cluster may be extracted based on the feature vectors.

[0046] The calculation unit 130 generates a representative characteristic table, as shown in FIG. 6, which stores representative characteristics for each cluster based on the calculated representative vector, and stores it in the user DB 110 (step S105), thereby completing the candidate classification process.

[0047] (Answer generation process) Next, the answer generation process executed by the generation device 100 will be described with reference to FIG.

[0048] The receiving unit 140 receives a question sentence and an instruction to generate an answer to the question sentence (Step S201). Specifically, the terminal 200 transmits the created question sentence and the instruction to generate an answer to the generation device 100 in accordance with the requester's operation.

[0049] When the receiving unit 140 receives the question text and the instruction to generate an answer (step S202; Yes), it transmits the received question text to the generation unit 150 and proceeds to step S203. On the other hand, when the receiving unit 140 does not receive the question text and the instruction to generate a mock answer (step S202; No), it returns to step S201 and receives the question text.

[0050] Next, the generating unit 150 repeatedly executes the processes of steps S204 to S206 for each cluster generated by the candidate classification process (step S203).

[0051] In step S204, the generation unit 150 generates a prompt that instructs the generation AI server 300 to generate answers by simulating the representative of each cluster, based on the question received in step S202 and the representative characteristics of each cluster. An example of the prompt is shown in FIG. 10. The example shown is a prompt for simulating the representative of cluster ID 1 to generate answers to questions 1 to 5. This prompt is generated based on the attributes and behavioral history of cluster ID 1 stored in the representative characteristics table shown in FIG. 6.

[0052] In step S205, the generation unit 150 obtains the answer. Specifically, the generation unit 150 transmits the prompt generated in step S204 to the generation AI server 300. The generation unit 150 obtains the answer generated by the generation AI server 300.

[0053] Next, in step S206, the generation unit 150 determines whether or not the processing of loop 1 has been executed for all clusters. If the generation unit 150 determines that there are unprocessed clusters, the generation unit 150 executes the processing of loop 1 for the unprocessed clusters. On the other hand, if the generation unit 150 determines in step S206 that the processing of loop 1 has been executed for all clusters included in the content management information, the process proceeds to step S207.

[0054] In step S207, the generation unit 150 outputs an answer table showing a list of the mock answers generated, and ends the process. An example of the answer table is shown in FIG. 11. As shown in the figure, the answer table is a table that stores cluster IDs and mock answers in association with each other. As shown in the example, in addition to the answers, the size of each cluster, i.e., the proportion of the number of candidates in each cluster to the total number of candidates, may also be stored. This allows the person requesting the survey to predict what types of answers to the question are likely to be collected most often. The answer table may also include the representative characteristics of each cluster and the question.

[0055] The person requesting the survey can view the mock answers and, if they decide that they need to revise the question, revise the question and run the answer generation process again. By repeatedly running the answer generation process, it is possible to create a more appropriate question that will yield an answer that meets the purpose of the question.

[0056] As described above, the generation device 100 classifies the candidates who will answer questions into multiple clusters based on their characteristic information and generates mock answers for the representatives of each cluster. This allows for obtaining a variety of answers to be used in advance to evaluate the effectiveness and appropriateness of the questions.

[0057] Furthermore, the generating device 100 classifies candidates based on the characteristic information of the actual candidates stored in the user DB 110, and provides the representative characteristics of the representatives representing each cluster to the generating AI server 300. This makes it possible to obtain a variety of answers that are closer to reality.

[0058] (Variation) In the above embodiment, the generation device 100 is described as receiving a question from the terminal 200 and generating a simulated answer. However, this is not limiting. The generation device 100 may also receive a question from a chatbot that obtains answers to a questionnaire through communication with a user and generate a simulated answer to the received question. In this case, the generation device 100 may be connected to a chatbot server, generate a prompt that generates an answer to a question sent from the chatbot, and transmit the prompt to the generation AI server 300. For example, the generation device 100 may transmit a prompt including a representative characteristic of one of multiple classified clusters to the generation AI server 300, execute a series of communications with the chatbot, and execute the same for the number of clusters. This allows, for example, simulations to verify the operation of a chatbot under development or to improve the performance of a chatbot currently in operation using a variety of response patterns without manual intervention. Note that the communication by the chatbot described above may be in the form of an interview. For example, instead of outputting common questions in advance, the chatbot server may output individual questions to each user in an interactive format to obtain answers by delving into the user's opinions (e.g., the user's feelings about the answers). That is, the generation device 100 may execute processing not only for questionnaire questions but also for interview questions.

[0059] The generation device 100 may further include a determination unit that determines whether the question sentence is appropriate. For example, the determination unit evaluates the diversity of answers and the relevance between the question and the answers based on the multiple answers for each cluster included in the answer table output in step S207, and determines whether the question sentence is appropriate based on the evaluation result. When evaluating the diversity of answers, the determination unit may, for example, convert each of the multiple generated answers into a vector, calculate the similarity between the answers using cosine similarity or the like, and evaluate the lower the similarity as higher the diversity, and vice versa. If the similarity is equal to or greater than a threshold, the determination unit may determine that the question sentence is inappropriate because the range of answers is limited, and output a determination result to that effect.

[0060] Furthermore, when evaluating the relevance between a question and an answer, the determination unit may, for example, vectorize the question and the generated answer, calculate the cosine similarity between the question and the answer, and calculate the average of the cosine similarities between the question sentence and each generated answer as the relevance score. Since the relevance score indicates how semantically related the generated answer is to the question, if the calculated relevance score is below a threshold, the determination unit may determine that the question sentence is inappropriate because the question is ambiguous and the intent of the question may not be conveyed, and output a determination result to that effect.

[0061] Furthermore, if the determination unit determines that the question is inappropriate, the generation unit 150 may send a prompt to the generation AI server 300 instructing the user to revise the question, thereby generating a question that takes the evaluation result into account. For example, if the question is evaluated as having low diversity, the generation unit 150 may generate a prompt such as, "Please revise the question so that it elicits different perspectives and various opinions." Alternatively, if the relevance between the question and the answer is evaluated as being low, the generation unit 150 may generate a prompt such as, "Please revise the question to make it clearer and more specific."

[0062] It should be noted that the generating device 100 according to the above embodiment can be realized using a normal computer, not a dedicated device. For example, the generating device 100 that executes the above-described processes may be configured by installing a program for executing any of the above-described processes on a computer from a recording medium storing the program on the computer. Furthermore, the generating device 100 may be configured by multiple computers operating in cooperation with each other.

[0063] Furthermore, when the above-mentioned functions are realized by sharing the functions between an OS (Operating System) and an application, or by cooperation between the OS and the application, only the parts other than the OS may be stored on the medium.

[0064] It is also possible to superimpose the program on a carrier wave and distribute it via a communication network. For example, the program may be distributed through an application store (App Store) or posted on a bulletin board system (BBS) on a communication network and distributed via the network. These programs may then be launched and run under the control of an operating system in the same way as other application programs, thereby enabling the above-described processing to be performed.

[0065] Furthermore, the information stored in the storage unit 14 may be collectively managed by a cloud server present on the network, and the generating device 100 may access the cloud server as needed to read and write information. In this case, the generating device 100 does not need to include the user DB 110. Furthermore, the candidate classification process and the answer generation process by the generating device 100 may be executed on the cloud using information stored in the cloud server.

[0066] Various aspects of the present disclosure are summarized below as appendices.

[0067] (Appendix 1) a classification unit that classifies candidates who can answer questions into a plurality of clusters based on characteristic information of the candidates; a calculation unit that calculates a representative characteristic that indicates a characteristic of a representative that represents each of the clusters based on the characteristic information of the candidates included in each of the classified clusters; a reception unit for receiving questions; a generation unit that causes a generation AI that simulates a representative having the representative characteristics obtained for each of the clusters to generate a simulated answer to the question sentence received by the reception unit; A generating device comprising:

[0068] (Appendix 2) The characteristic information includes attributes and behavioral history of the candidate. 10. The generating device of claim 1.

[0069] (Appendix 3) the calculation unit calculates the representative characteristic for each cluster based on the characteristic information of a candidate who is closest to a center of gravity of the cluster among the candidates who belong to the classified cluster; 3. The generating device of claim 1 or 2.

[0070] (Appendix 4) the calculation unit calculates, for each of the classified clusters, an average value of the characteristic information in the cluster as the representative characteristic. 3. The generating device of claim 1 or 2.

[0071] (Appendix 5) The generation unit generates a prompt that instructs the generation AI to generate an answer to the question by simulating a representative having the representative characteristics based on the question received by the reception unit and the representative characteristics calculated by the calculation unit, and transmits the generated prompt to the generation AI. 5. The generating device of any one of claims 1 to 4.

[0072] (Appendix 6) the generation unit outputs the generated answer in association with a size of a cluster related to the answer. 6. The generating device of any one of claims 1 to 5.

[0073] (Appendix 7) The computer A step of classifying candidates who can answer the question into a plurality of clusters based on characteristic information of the candidates; A step of determining a representative characteristic that indicates a characteristic of a representative that represents each of the clusters based on the characteristic information of the candidates included in each of the classified clusters; accepting a question; A step of causing a generation AI that simulates a representative having the representative characteristics obtained for each of the clusters to generate a simulated answer to the received question sentence; A generation method that performs the following.

[0074] (Appendix 8) On the computer, A process of classifying candidates who can answer the questions into a plurality of clusters based on characteristic information of the candidates; A process of determining a representative characteristic that indicates a characteristic of a representative who represents each of the clusters based on the characteristic information of the candidates included in each of the classified clusters; A process for accepting a question; A process of causing a generation AI that simulates a representative having the representative characteristics obtained for each of the clusters to generate a simulated answer to the received question; A program that executes the following.

[0075] The present disclosure allows various embodiments and modifications without departing from the broad spirit and scope of the present disclosure. Furthermore, the above-described embodiments are intended to illustrate the present disclosure and do not limit the scope of the present disclosure. That is, the scope of the present disclosure is defined by the claims, not the embodiments. Various modifications made within the scope of the claims and the meaning of equivalent disclosures are considered to be within the scope of the present disclosure. [Industrial Applicability]

[0076] The present invention can be suitably employed in a generation device, a generation method, and a program that are capable of obtaining a variety of answers for evaluating the effectiveness and appropriateness of a question in advance. [Explanation of symbols]

[0077] 100 generation device, 200 terminal, 300 generation AI server, 400 communication network, 110 user DB, 120 classification unit, 130 calculation unit, 140 reception unit, 150 generation unit, 11 CPU, 12 RAM, 13 ROM, 14 storage unit, 15 input unit, 16 display unit, 17 communication unit, 99 internal bus

Claims

1. a classification unit that classifies candidates who can answer questions into a plurality of clusters based on characteristic information of the candidates; a calculation unit that calculates a representative characteristic that indicates a characteristic of a representative that represents each of the clusters based on the characteristic information of the candidates included in each of the classified clusters; a reception unit for receiving questions; a generation unit that causes a generation AI that simulates a representative having the representative characteristics obtained for each of the clusters to generate a simulated answer to the question sentence received by the reception unit; A generating device comprising:

2. The characteristic information includes attributes and behavioral history of the candidate. The generating device of claim 1 .

3. the calculation unit calculates the representative characteristic for each cluster based on the characteristic information of a candidate who is closest to a center of gravity of the cluster among the candidates who belong to the classified cluster; 3. The generating device according to claim 1 or 2.

4. the calculation unit calculates, for each of the classified clusters, an average value of the characteristic information in the cluster as the representative characteristic.

3. The generating device according to claim 1 or 2.

5. The generation unit generates a prompt that instructs the generation AI to generate an answer to the question by simulating a representative having the representative characteristics based on the question received by the reception unit and the representative characteristics calculated by the calculation unit, and transmits the generated prompt to the generation AI.

3. The generating device according to claim 1 or 2.

6. the generation unit outputs the generated answer in association with a size of a cluster related to the answer.

3. The generating device according to claim 1 or 2.

7. The computer A step of classifying candidates who can answer the question into a plurality of clusters based on characteristic information of the candidates; A step of determining a representative characteristic that indicates a characteristic of a representative that represents each of the clusters based on the characteristic information of the candidates included in each of the classified clusters; accepting a question; A step of causing a generation AI that simulates a representative having the representative characteristics obtained for each of the clusters to generate a simulated answer to the received question sentence; A generation method that performs the following.

8. On the computer, A process of classifying candidates who can answer the questions into a plurality of clusters based on characteristic information of the candidates; A process of determining a representative characteristic that indicates a characteristic of a representative who represents each of the clusters based on the characteristic information of the candidates included in each of the classified clusters; A process for accepting a question; A process of causing a generation AI that simulates a representative having the representative characteristics obtained for each of the clusters to generate a simulated answer to the received question; A program that executes the following.

Images