Personality Evaluation System

Abstract

Provided is a personality evaluation system S that can appropriately evaluate the personality of user U. 【Solution means】The personality evaluation system S of the present invention includes a service management device 1 that provides and manages a conversation service for the user terminal 3, and a generative AI 2 that generates a virtual imitation personality C corresponding to the user U who uses the user terminal 3. The generative AI 2 includes an imitation personality generation unit 21 that generates an imitation personality C reflecting the personality of the user U based on the conversation information of the user U, a skill correction unit 22 that corrects the imitation personality C corresponding to the skills of the user U, and a value correction unit 23 that corrects the imitation personality C corresponding to the threshold value of the values of the user U obtained from the conversation information of the user U.

Description

Technical Field

[0001] The present invention relates to a personality evaluation system for evaluating a user's personality.

Background Art

[0002] Conventionally, for example, a so-called chat service that performs real-time character communication or the like using an information processing terminal such as a smartphone, a tablet, or a computer has been provided. Examples of chat modes include one-on-one user chats between users, in-community chats among a plurality of users belonging to a specific community, or chats between communities. This type of chat service technology is disclosed in, for example, Patent Document 1.

[0003] In a chat service, in general, in order to start a chat, users are invited to participate through a predetermined approval process such as friend registration or community registration, and in this way, the number of users is increased.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, in the conventional case, since users increase without the involvement of service providers, there is a concern that, for example, acts of blaming, slandering, or defaming others may occur, either consciously or unconsciously, by an unspecified number of users, or that users who dislike this type of behavior may stop using the chat service, resulting in the chat service itself becoming sparse.

[0006] Possible countermeasures in this case include the service provider constantly checking and monitoring the content of chat messages to prevent the above problems. However, this not only requires a large amount of man-hours and is time-consuming, but also increases costs, and it is not possible to prevent everything.

[0007] If, hypothetically, the personalities of all users who use the chat service, such as their characters and temperaments, are grasped, it is understood that it becomes possible to prevent troubles between users, for example, by prompting attention to users who are likely to cause troubles. Also, it is understood that it becomes possible to promote the participation of users who are good at livening up the atmosphere in communities that are becoming sparsely populated and activate the relevant communities. That is, if one has an understanding of the personalities of all users who use the chat service, it is considered that it is possible to achieve both the smooth operation and the widespread expansion of the chat service without the service provider constantly checking and monitoring the content of chat messages.

Means for Solving the Problem

[0008] The present invention has been made in view of the above current situation, and aims to provide a personality evaluation system that appropriately evaluates the personalities of users as a technical problem.

[0009] The personality evaluation system according to the present invention includes a service management device that provides and manages a conversation service for a user terminal, and an imitation personality generation device that generates a virtual imitation personality corresponding to a user who uses the user terminal. The imitation personality generation device includes an imitation personality generation unit that generates an imitation personality reflecting the personality of the user based on the conversation information of the user, a skill correction unit that corrects the imitation personality corresponding to the skills possessed by the user, and a value correction unit that corrects the imitation personality corresponding to the value threshold of the user obtained from the conversation information of the user. and the imitation personality generation device performs a personality diagnosis in a free conversation format for each of the user and the imitation personality, and the service management device compares the personality diagnosis information of the user and the personality diagnosis information of the imitation personality, and determines the imitation accuracy of the imitation personality from the similarity is what it is.

[0011] In the personality evaluation system of the present invention, the service management device may have a large number of the imitated personalities for each user, and cause the actions of the arbitrary user obtained from the conversation information of the arbitrary user to be evaluated by the large number of imitated personalities, and set the evaluation result as the characteristic waveform of the arbitrary user.

[0012] In the personality evaluation system of the present invention, the service management device may compare the characteristic waveform of the arbitrary user with the characteristic waveform of another arbitrary user, and determine whether or not the two users match based on the similarity.

[0013] In the personality evaluation system of the present invention, the service management device may set an aggregate of the characteristic waveforms of a plurality of arbitrary users as the characteristic waveform of a specific community, compare the characteristic waveform of the arbitrary user with the characteristic waveform of the specific community, and determine whether or not the arbitrary user matches the specific community based on the similarity.

[0014] In the personality evaluation system of the present invention, the service management device may set an aggregate of the characteristic waveforms of a plurality of arbitrary users as the characteristic waveform of a specific community, compare the characteristic waveform of the specific community with the characteristic waveform of another specific community, and determine whether or not the two specific communities match based on the similarity.

Effect of the Invention

[0015] According to the present invention, the personality of each user using the conversation service can be easily grasped from the corresponding imitated personality. Therefore, for example, restrictions can be imposed on the use of the conversation service for users who are highly likely to cause trouble, thereby preventing trouble between users. In addition, it is possible to promote the participation of users who are good at livening up the atmosphere in a community that is becoming sparsely populated, and activate the community. In short, it is possible to achieve both smooth operation and widespread expansion of the conversation service without the service provider constantly checking and monitoring the conversation content.

Brief Description of the Drawings

[0016]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Mode for Carrying Out the Invention

[0017] Next, embodiments embodying the present invention will be described with reference to the drawings. Although the drawings show preferred embodiments, it can be implemented in many different forms and is not limited to the embodiments described in this specification.

[0018] First, with reference to FIG. 1 and the like, an overview of the personality evaluation system S (hereinafter simply referred to as "system S") according to the embodiment will be described. The system S evaluates the personality of the user U using the generation-based AI 2 (generation-based artificial intelligence) as an imitation personality generation device based on the conversation content of the user U in a chat service, which is an example of a conversation service.

[0019] As shown in FIG. 1, the system S includes a service management device 1 operated by the service provider of the system S, a generation-based AI 2, and a plurality of user terminals 3, etc.

[0020] The service management device 1 is composed of one or more computers and functions as, for example, a web server on the communication network 4. The service management device 1 in the embodiment is configured to provide and manage a chat service for each user terminal 3. The service management device 1 is configured to be communicably connected to each user terminal 3 via the communication network 4 and is capable of exchanging various information with each user terminal 3.

[0021] The generative AI 2 utilizes a method including known AI programs, AI libraries, or AI platforms typified by machine learning and deep learning. As AI technologies corresponding to text data, voice data, or video data, for example, known models and methods such as hidden Markov models, convolutional neural networks (CNNs), recurrent neural networks (RNNs), or LSTMs (Long Short-Term Memory) are used.

[0022] In the embodiment, the generative AI 2 uses a learning model based on a large language model (LLM) such as ChatGPT. The generative AI 2 is configured to be communicably connected to the service management device 1 via an API (Application Programming Interface).

[0023] The generative AI 2 extracts emotions such as joy, anger, sorrow, and happiness, positive behaviors such as gratitude and praise, and negative behaviors such as slander and libel based on the user behavior information 7 of each user U. Then, the generative AI 2 can perform a personality evaluation of the user U using the extracted results as an index. For example, a user U who uses the chat app 5 less frequently or engages in negative behaviors such as slandering other users U will receive a lower evaluation. A user U who enlivens the atmosphere, performs positive behaviors such as gratitude and praise, or uses language that is well-liked by other users U will receive a higher evaluation. Note that the user behavior information 7 includes, for example, various texts such as chats, images, and files, and may include not only characters but also voice, images, or videos.

[0024] The generative AI 2 includes an imitation personality generation unit 21 that generates an imitation personality C reflecting the personality of the user U based on the user behavior information 7 of the user U, a skill correction unit 22 that corrects the imitation personality C corresponding to the skills of the user U, a value correction unit 23 that corrects the imitation personality C corresponding to the threshold value of the values of the user U obtained from the user behavior information 7 of the user U, and the like.

[0025] The imitation personality generation unit 21 generates each imitation personality C (the program constituting this) by analyzing the user behavior information 7 of each user U collected and stored in the user behavior information DB 19 of the storage unit 13 described later. Each imitation personality C is constructed by a computer program and behaves as a virtual human being just like an actual human being, having a personality and emotions imitating the corresponding user U. The emotions of each imitation personality C are set to change in the same way as the emotions of the corresponding user U. In other words, each imitation personality C is set to simulate (or emulate) the emotions of the corresponding user U.

[0026] The imitation personality C is an artificial intelligence (AI) that imitates the personality, emotions, and emotional changes of the user U. Specifically, it includes a program that executes an algorithm and data processed by the program. The imitation personality C has algorithms corresponding to a learning function that learns about the user U based on the user behavior information 7, an emotion detection function that estimates (detects) the emotions of the user U, a conversation execution function that conducts conversations with the user U, and the like. Each imitation personality C is executed and processed by the imitation personality execution unit 17 (details will be described later) of the information processing unit 12 in the service management device 1.

[0027] The skill correction unit 22 executes a process of correcting the tendency of the imitation personality C for the user U based on the skills (such as occupation, etc.) of each user U. The skills of the user U mean advanced abilities cultivated through training and learning. In the embodiment, the skills of the user U are inferred from the occupation of the user U, but the skills may also be recognized from other abilities of the user U.

[0028] Based on the user behavior information 7 of each user U, the value correction unit 23 analyzes the thresholds of each judgment (value concept) such as good and evil, likes and dislikes, gains and losses, interests, pain and pleasure, difficulty and ease, feasibility, pleasure and displeasure, and right and wrong that the user U has, and executes a process of setting the analysis result as the threshold of each judgment in the corresponding imitated personality C.

[0029] Each user terminal 3 is an information processing terminal by which the user U who owns it receives the chat service. Examples of the user terminal 3 include a smartphone, a tablet, or a computer. Each user terminal 3 is configured to be communicably connected to the service management device 1 via the communication network 4, and can exchange various information with the service management device 1.

[0030] Each user terminal 3 can communicate with different terminals 3 via the communication network 4 through the service management device 1. When using the chat service for the first time, the user terminal 3 is used to access the service management device 1, and the chat application 5 is downloaded and installed on the user terminal 3.

[0031] User information including the personal data of the user U and the like is stored in each user terminal 3. Examples of the user information include name, gender, date of birth (age), occupation, mobile phone number, and the user ID 6 (see FIG. 2) of the chat service and community ID. Note that the communication network 4 is composed of, for example, the Internet, a mobile communication network, and a base station, etc., but there is no limitation on the type of communication protocol, nor on the type and scale of the network.

[0032] FIG. 2 is a conceptual diagram showing the hardware configuration of the service management device 1. As shown in FIG. 2, the service management device 1 includes a communication unit 11, an information processing unit 12, a storage unit 13, and the like. The communication unit 11, the information processing unit 12, the storage unit 13, and the like are respectively connected to each other via a system bus 14.

[0033] The communication unit 11 is, for example, a NIC for Ethernet or various communication modules for communication such as wired LAN and wireless LAN, and has a function of communicatively connecting to the communication network 4.

[0034] The information processing unit 12 includes a RAM, a ROM, a CPU (not shown), etc. The information processing unit 12, according to a chat management program etc. stored in the storage unit 13, executes, for example, a process of acquiring chat information, which is conversation information, from each user terminal 3 and recording it in the storage unit 13, and a process of transmitting it to other user terminals 3 belonging to a community etc.

[0035] The information processing unit 12 has a personality similarity calculation unit 15, a waveform similarity calculation unit 16, an imitation personality execution unit 17, a value waveform generation unit 24, etc. The personality similarity calculation unit 15 executes a process of comparing the personality diagnosis information of the user U and the imitation personality C obtained from the conversation with the generation system AI2 and obtaining a personality similarity indicating the degree of match. Although details will be described later, the value waveform generation unit 24 executes a process of obtaining value waveforms G+(U or SC), G-(U or SC) as characteristic waveforms indicating the cumulative results of virtues and landmines for each imitation personality C regarding each user U, a specific community SC, etc. Also, the waveform similarity calculation unit 16 executes a process of comparing the value waveforms G+(U), G-(U) of each user U and the value waveforms G+(SC), G-(SC) of a specific community SC and determining the degree of match. The imitation personality execution unit 17 is responsible for the execution process of the imitation personality C, which is a program including an artificial intelligence algorithm.

[0036] Here, the personality similarity is an index indicating the degree of similarity between the personality diagnosis information of the user U and the imitation personality C. As the personality similarity, for example, the cosine similarity between the feature vector of the personality diagnosis information of the imitation personality C and the feature vector of the personality diagnosis information of the user U may be used. The cosine similarity evaluates the degree of similarity between information based on the closeness (angle) of the vectors of the information.

[0037] More specifically, for example, the personality similarity calculation unit 15 first calculates the number a of all elements of the feature vector where the values of both personality diagnosis information are 1. Then, the personality similarity calculation unit 15 calculates the number b of all elements of the feature vector where the value of either one of the personality diagnosis information is 1. And the personality similarity calculation unit 15 divides the calculated number a by the number b to calculate the cosine similarity (a / b) between the feature vector of the personality diagnosis information of the mimicking personality C and the feature vector of the personality diagnosis information of the user U. In this case, the maximum value of the personality similarity is "1", and the minimum value is "0".

[0038] In addition, the waveform similarity is an index indicating the degree of similarity between the value waveform G+(U), G-(U) of the user U, the degree of similarity between the value waveform G+(U), G-(U) of the user U and the value waveform G+(SC), G-(SC) of a specific community SC, or the degree of similarity between the value waveform G+(SC), G-(SC) of a certain specific community SC and the value waveform G+(SC), G-(SC) of another specific community SC. As the waveform similarity, the cosine similarity between the feature vectors of the value waveforms may be used.

[0039] Note that the method for calculating the personality similarity and the waveform similarity is not limited to the cosine similarity, and other methods such as using another calculation function or using other methods such as template matching and pattern matching may be used. That is, any method may be applied to calculate the personality similarity and the waveform similarity.

[0040] The storage unit 13 is a non-volatile memory such as an HDD (Hard Disc Drive), a flash memory (SSD: Solid State Drive), or other solid-state memories, and stores an operating system, a chat management program, etc. The storage unit 13 stores and manages the user ID6 of the user U who uses the chat service, the community ID of the chat community, etc., and records the user behavior information 7 as the conversation information acquired from each user terminal 3 in association with the corresponding user ID6.

[0041] As a database containing this type of data, in the storage unit 13 of the embodiment, a user information database 18 (hereinafter referred to as "user information DB18"), a user behavior information database 19 (hereinafter referred to as "user behavior information DB19"), an imitation personality database 20 (hereinafter referred to as "imitation personality DB20"), etc. are constructed. Note that each of the DBs 18 to 20 may be provided in a network server (not shown) on a communication network 4 separate from the service management device 1, or may be provided in a local server (not shown) separately connected to the service management device 1.

[0042] It is desirable to store (store) various information stored and managed in the storage unit 13 on the blockchain 8 (see FIG. 1). It is preferable to store at least various information stored in the user information DB18 in the storage unit 13 on the blockchain 8. The blockchain 8 (decentralized ledger) is based on a P2P network composed of a plurality of computer-related devices (nodes) connected to the communication network 4, and is constructed by, for example, a decentralized application and a smart contract provided by Ethereum, Polygon, etc.

[0043] When storing various information on the blockchain 8, it is preferable to configure it such that each user terminal 3 generates a transaction to the blockchain 8, and when the transaction is approved on the blockchain 8, the various information stored on the blockchain 8 is updated. The service management device 1 may function as a node constituting the blockchain 8. Note that as the blockchain 8, a public blockchain such as Ethereum or Polygon may be applied, or other private blockchains may be applied.

[0044] The user information DB18 is a database that registers various information of user U for whom the user account of system S is issued. In the user information DB18, a part of the user information described above, that is, the user ID6 for user identification, password, the occupation of user U, the community ID indicating the chat community to which user U belongs, etc. are registered in association with each user terminal 3.

[0045] The user behavior information DB19 is a database that collects and accumulates a large number of user behavior information 7 such as various texts, images, and files of messages (conversations) during the use of the chat service obtained from each user terminal 3, in association with each user ID6. The user behavior information DB19 includes a table that summarizes, as a history, the user behavior information 7 obtained from each user terminal 3 for each user ID6 that identifies an individual user U, at appropriate time intervals (for example, every 24 hours (one day) or every hour). The imitation personality DB20 is a database that stores and remembers a plurality of imitation personalities C generated by the generative AI2, in association with each user ID6.

[0046] In the embodiment, each imitation personality C (the program constituting this) is generated by the imitation personality generation unit 21 (see FIG. 1) of the generative AI2 analyzing the user behavior information 7 of each user U collected and accumulated in the user behavior information DB19 of the storage unit 13.

[0047] In this case, the imitation personality generation unit 21 of the generative AI2 refers to, for example, the Big Five theory in psychology, etc., extracts emotions such as joy, anger, sorrow, and happiness, positive behaviors such as gratitude and praise, and negative behaviors such as slander and libel based on the user behavior information 7 of each user U, and analyzes the personality and emotions (so-called personality) of the user U. In the Big Five theory, the personality of user U is classified into five characteristics. The five characteristics evaluated by the generative AI2 (imitation personality generation unit 21) include agreeableness, conscientiousness, extraversion, neuroticism, and openness.

[0048] For example, cooperativeness means the characteristic of trying to adjust social harmony and the individual. A user U with a high evaluation of cooperativeness is generally understood to be considerate, kind, generous, and trustworthy, and has a personality that tries to compromise the interests of others and himself / herself. Honesty means self-control, and a user U with a high evaluation of honesty is understood to have the characteristic of acting honestly and pursuing achievements in response to external expectations. Extroversion is the characteristic of obtaining energy from various activities or the external environment. A user U with a high evaluation of extroversion is understood to like to communicate with other users U and is enthusiastic and active.

[0049] Neurotic tendency means the tendency to easily feel negative emotions such as anger, anxiety, or depression. A user U with a high evaluation of neurotic tendency is understood to be prone to emotional reactions, vulnerable to stress, and tend to change depending on the way of expressing emotions. Openness is an evaluation of art, emotions, adventure, imagination, curiosity, and various experiences, etc. A user U with a high evaluation of openness is understood to be rich in intellectual curiosity, sensitive to beauty, and have a personality that happily tries new things.

[0050] The imitation personality generation unit 21 of the generative AI 2 assigns a tendency from 0% to 100% to the five characteristics, indexes (for example, scores, etc.) each characteristic of the user U based on such a tendency, and converts them as each characteristic of the corresponding imitation personality C.

[0051] Then, the skill correction unit 22 of the generative AI 2 corrects the tendency of each characteristic based on the skill (occupation in the embodiment) of each user U, reflects the corrected tendency of each characteristic as an index (for example, score, etc.) of each characteristic of the user U, and converts them as the corrected each characteristic of the corresponding imitation personality C.

[0052] Users U (imitation personalities C) with similar distributions of the indexes (for example, scores, etc.) obtained by the skill correction unit 22 of the generative AI 2 are evaluated as having similar (close) personalities. Users U (imitation personalities C) with completely different distributions of the indexes (for example, scores, etc.) are evaluated as having different (distant) personalities.

[0053] Note that as an analysis method for personality and the like, it is not limited to the Big Five theory, and other analysis methods such as MBTI (Myers-Briggs Type Indicator), StrengthsFinder (registered trademark), and others may also be used.

[0054] Furthermore, the value correction unit 23 of the generative AI 2 analyzes the thresholds of each judgment (values) such as good and evil, likes and dislikes, gains and losses, interests and harms, pain and pleasure, difficulty and ease, approval and disapproval, pleasure and displeasure, and right and wrong that the user U has based on the user behavior information 7 of each user U, standardizes them (for example, converts them into scores, etc.), and converts them into the thresholds of each judgment in the corresponding mimetic personality C.

[0055] Also in this case, users U (mimetic personalities C) with similar distributions of the indicators (for example, scores, etc.) obtained by the value correction unit 23 of the generative AI 2 are evaluated as having similar (close) judgment senses (values). Users U (mimetic personalities C) with completely different distributions of the indicators (for example, scores, etc.) are evaluated as having different (distant) judgment senses (values).

[0056] Each mimetic personality C has different value thresholds (which can also be said to be judgment scales and judgment criteria) from each other. That is, the thinking and judgment criteria of each mimetic personality C are made to be individually different like those of ordinary people. This is adopted in view of the fact that the method of quantifying human thinking and judgment criteria with a single criterion lacks reliability.

[0057] For example, due to differences in the degree of tolerance for "being loose with time" or "vulgar language", etc., while one mimetic personality C gives a zero evaluation (does not regard it as a problem), another mimetic personality C gives a negative evaluation or multiple negative evaluations. That is, for each behavioral item of each user U, whether it is a positive evaluation, a negative evaluation, or a zero evaluation is entrusted to each mimetic personality C. Generally speaking, a positive evaluation is considered when the mimetic personality C makes a favorable judgment like positive behaviors such as gratitude and praise, and a negative evaluation is considered when the mimetic personality C makes an unfavorable judgment like negative behaviors such as slander and libel.

[0058] Note that the speech and action items for evaluating each user U are recalled by the generative AI 2 or any imitation personality C based on the user speech and action information 7 of each user U, and the recalled speech and action items are set to be evaluated and judged by all imitation personalities C as common matters. In the embodiment, positive evaluation may be expressed as "virtue", the accumulation of positive evaluation as "accumulating virtue", negative evaluation as "landmine", and the accumulation of negative evaluation as "continuously stepping on landmines".

[0059] In the system S of the embodiment, each imitation personality C (the program constituting this) generated by the generative AI 2 is stored in the imitation personality DB 20 of the recording unit 13 in the service management device 1. And although details will be described later, the service management device 1 (imitation personality execution unit 17) executes the program constituting each imitation personality C. That is, by the service management device 1 (imitation personality execution unit 17) executing the program constituting each imitation personality C, each imitation personality C operates.

[0060] In this case, the service management device 1 detects and collects the user speech and action information 7 of each user U at a predetermined timing. The user speech and action information 7 of each user U that has been added or updated is also stored in the user speech and action information DB 19 in association with each user ID 6. The generative AI 2 (imitation personality generation unit 21) analyzes the user speech and action information 7 of each user U that has been added or updated and causes the corresponding imitation personality C to learn, or the imitation personality execution unit 17 of the service management device 1 analyzes the user speech and action information 7 of each user U that has been added or updated and causes the corresponding imitation personality C to learn.

[0061] Through such learning, each imitation personality C can imitate the personality, emotions, and judgment criteria (thresholds) of the corresponding user U with higher precision, and can acquire an algorithm for reproducing the personality, emotions, and judgment criteria (thresholds) with high precision specialized for the user U. The algorithm of each imitation personality C is improved and modified each time through the accumulation of learning. The service management device 1 updates each imitation personality C as the learning of each imitation personality C progresses and stores it in the imitation personality DB 20.

[0062] Figure 3 is a conceptual diagram showing the hardware configuration of the user terminal 3. As shown in Figure 3, the user terminal 3 has a CPU 31, a memory 32, a storage 33, a communication IF 34, an input device 35, a display unit 36, an input / output IF 37, etc. These are respectively connected to each other via a system bus 38. The CPU 31 performs various calculations according to a program and performs processing to realize the functions of the user terminal 3. Note that the number of CPUs 31 may be single or plural. When using a plurality of CPUs 31, these CPUs 31 may execute processing simultaneously or execute sequential processing.

[0063] The memory 32 functions as a work area when the CPU 31 executes processing. The memory 32 includes, for example, a RAM and a ROM. The storage 33 stores various programs and data, and includes, for example, an SSD and / or an HDD. The storage 33 stores a program for realizing the functions of the user terminal 3, such as the chat application 5 provided by the service management device 1.

[0064] The communication IF 34 communicates with other devices according to a predetermined communication standard (for example, Ethernet). The communication IF 34 includes, for example, a NIC. The input device 35 inputs information according to the operation of the user U. The input device 35 includes at least one of, for example, a touch sensor, a keyboard, a keypad, a mouse, and a microphone.

[0065] The display unit 36 displays various information. The display unit 36 includes, for example, an LCD. The display unit 36 may be integrated with a touch sensor to be configured as a touch screen. The input / output IF 37 is an interface for connecting the input device 35, a camera, etc.

[0066] Next, an example of generating an imitation personality Cn (the subscript n is an integer of 1 or more) corresponding to an arbitrary user Un among the operations of the personality evaluation system S according to the embodiment will be described with reference to FIG. 4.

[0067] As shown in FIG. 4, when user Un starts the chat application 5 on his / her user terminal 3 and starts actions such as chatting with other user U (actions from which information can be obtained) (step S01: YES), the service management device 1 collects and accumulates user action information 7n such as messages from the user terminal 3n of user Un in the user action information DB 19 in association with the user ID 6n (step S02).

[0068] Next, when a predetermined amount of user action information 7n is collected (step S03: YES), the service management device 1 sends a generation request for the imitation personality Cn including the user ID 6n and the user action information 7n to the generative AI 2 via the API (step S04).

[0069] Upon receiving the generation request for the imitation personality Cn from the service management device 1, the generative AI 2 causes the imitation personality generation unit 21 to standardize (for example, score) each characteristic of user Un through analysis based on the user action information 7n in response to the imitation personality Cn generation request, and generates the imitation personality Cn (program) (step S05). Next, the skill correction unit 22 of the generative AI 2 reflects the correction based on the occupation of user Un in the imitation personality Cn (step S06). Then, based on the user action information 7n, the generative AI 2 causes the value judgment correction unit 23 to analyze and standardize (for example, score) the threshold of each judgment of user Un, and reflects it in the imitation personality Cn (step S07).

[0070] When the generation of the imitation personality Cn by the generative AI 2 is completed, a generation completion notification of the imitation personality Cn including the imitation personality Cn (program) is sent from the generative AI 2 to the service management device 1 via the API (step S08). Next, the service management device 1 that has received the generation completion notification of the imitation personality Cn from the generative AI 2 associates the imitation personality Cn with the user ID 6n and registers and stores it in the imitation personality DB 20 of the storage unit 13 (step S09).

[0071] According to the above control, there is a service management device 1 that provides and manages a conversation service for the user terminal 3, and an imitation personality generation device 2 (generative AI) that generates a virtual imitation personality C corresponding to the user U who uses the user terminal 3. The imitation personality generation device 2 includes an imitation personality generation unit 21 that generates an imitation personality C that reflects the personality of the user U based on the conversation information of the user U, a skill correction unit 22 that corrects the imitation personality C corresponding to the skills of the user U, and a value correction unit 23 that corrects the imitation personality C corresponding to the threshold value of the values of the user U obtained from the conversation information of the user U. Therefore, the personality of each user U who uses the conversation service can be easily grasped from the corresponding imitation personality C. For this reason, for example, restrictions can be imposed on the use of the conversation service for a user U who is highly likely to cause trouble, thereby preventing trouble between users U. In addition, it is possible to promote the participation of a user U who is good at livening up a community that is becoming sparse, and activate the community. In short, it is possible to achieve both the smooth operation and the spread of the conversation service without the service provider constantly checking and monitoring the conversation content.

[0072] Next, an example of improving the accuracy of the imitation personality Cn corresponding to an arbitrary user Un in the operation of the personality evaluation system S according to the embodiment will be described with reference to FIG. 5. In this case, it is assumed that the imitation personality Cn corresponding to the user Un is already stored in the storage unit 13 (imitation personality DB 20) of the service management device 1.

[0073] As shown in FIG. 5, while the user Un is performing actions such as chatting with other users U (actions from which information can be obtained) (step T01: YES), the service management device 1 determines whether the result of highly accurately imitating the personality diagnosis of the user Un has been executed (step T02). If the personality diagnosis of the user Un has not been executed (step T02: NO), the service management device 1 sends a request to the generative AI 2 via the API to execute a personality diagnosis in a free conversation format for the user Un (step T03).

[0074] Upon receiving a personality diagnosis execution request from the service management device 1, the generative AI 2 generates a large number of questions in a free conversation format and conducts a conversation including questions with the user Un in response to the personality diagnosis execution request (step T04).

[0075] The personality diagnosis by the generative AI 2 is not in the form of "the user Un selects from several options" (static personality diagnosis) as in the past, but is conducted in a form that encourages the user Un to make a smooth judgment by chatting (having a conversation) between the user Un and the generative AI 2. This is because in static personality diagnosis, since the questions are preset such that "the user Un selects from several options", there is room for consideration in terms of the user Un's answer being inconsistent or lacking in accuracy.

[0076] In contrast, for example, the generative AI 2 continuously conducts the chat without interruption and generates questions according to the content and flow while adapting to the response of the user Un. That is, the generative AI 2 behaves dynamically according to the response of the user Un and executes a personality diagnosis (dynamic personality diagnosis) for the user Un. If the generative AI 2 executes such a dynamic personality diagnosis, the generative AI 2 can obtain all the information necessary for the personality diagnosis of the user Un from the natural conversation flow, and the information obtained can also accurately reflect the personality of the user Un.

[0077] The personality diagnosis used in the embodiment is, for example, MBTI (Myers-Briggs Type Indicator). Since the diagnostic method of MBTI is already known, a detailed description thereof will be omitted. Note that the personality diagnosis method used here is not limited to MBTI, and may be the Big Five theory, StrengthsFinder (registered trademark), or other personality diagnosis methods mentioned above.

[0078] When the personality diagnosis of user Un by the generative AI 2 is completed, a personality diagnosis completion notification including the personality diagnosis information of user Un is sent from the generative AI 2 to the service management device 1 via the API (step T05). Next, the service management device 1 that has received the personality diagnosis completion notification from the generative AI 2 associates the personality diagnosis information of user Un with the user ID 6n and registers and stores it in the storage unit 13 (step T06).

[0079] Next, the service management device 1 activates the imitation personality Cn corresponding to user Un and sends a request to the generative AI 2 via the API to execute a personality diagnosis in the form of a free conversation for the imitation personality Cn (step T07). In this case, the execution process of the imitation personality Cn is carried out by the imitation personality execution unit 17. The generative AI 2 that has received the personality diagnosis execution request from the service management device 1 generates a large number of questions in the form of a free conversation in response to the personality diagnosis execution request and executes a conversation including the questions with the imitation personality Cn (step T08). In step T08, the generative AI 2 executes a dynamic personality diagnosis for the imitation personality Cn in the same manner as in step T04.

[0080] When the personality diagnosis of the imitation personality Cn by the generative AI 2 is completed, a personality diagnosis completion notification including the personality diagnosis information of the imitation personality Cn is sent from the generative AI 2 to the service management device 1 via the API (step T09). Next, the service management device 1 (personality similarity calculation unit 15) that has received the personality diagnosis completion notification from the generative AI 2 associates the personality diagnosis information of the imitation personality Cn with the user ID 6n and registers and stores it in the storage unit 13 (step T10).

[0081] Next, the service management device 1 (personality similarity calculation unit 15) reads out the personality diagnosis information of user Un and the personality diagnosis information of the imitation personality Cn from the storage unit 13, compares them, and calculates the personality similarity of the imitation personality C with respect to user U (step T11). If the personality similarity of the imitation personality C with respect to user U exceeds a predetermined threshold (step T11: YES), it can be said that the imitation personality Cn highly accurately imitates the personality, emotions, and judgment criteria (thresholds) of user Un and has acquired an algorithm for highly accurately reproducing the personality, emotions, and judgment criteria (thresholds) specialized for the user Un. Therefore, assuming that the imitation personality Cn is completed, the high-precision processing of the imitation personality Cn is terminated.

[0082] If the personality similarity of the imitation personality C with respect to user U is below the predetermined threshold (step T11: NO), the service management device 1 transmits an update request for the imitation personality Cn including the user behavior information 7n of the additional or updated user Un and the user ID 6n to the generative AI 2 via the API (step T12).

[0083] Upon receiving the update request for the imitation personality Cn from the service management device 1, the generative AI 2 (imitation personality generation unit 21) standardizes (for example, scores) each characteristic of the user Un through analysis based on the user behavior information 7n in response to the update request for the imitation personality Cn, and updates the imitation personality Cn (the program constituting this) (step T13). Then, the skill correction unit 22 of the generative AI 2 reflects the correction based on the skill (for example, occupation, etc.) of the user Un in the imitation personality Cn (step T14). Furthermore, the value view correction unit 23 of the generative AI 2 analyzes and standardizes (for example, scores) the thresholds of each judgment of the user Un based on the additional or updated user behavior information 7n, and reflects it in the imitation personality Cn (step T15).

[0084] When the update of the imitation personality Cn by the generative AI2 is completed, a notification of the completion of the update of the imitation personality Cn (program) including the imitation personality Cn is sent from the generative AI2 to the service management device 1 via the API (step T16). Next, the service management device 1 that has received the notification of the completion of the update of the imitation personality Cn from the generative AI2 associates the imitation personality Cn with the user ID6n and registers and stores it in the imitation personality DB20 of the storage unit 13 (step T17).

[0085] After that, the service management device 1 returns to step T07 and sends a request to the generative AI2 via the API to execute a personality diagnosis in the form of free conversation for the updated imitation personality Cn. Subsequently, the processes of steps T08 to T16 are looped and executed until the personality similarity of the updated imitation personality Cn to the user Un exceeds a predetermined threshold value.

[0086] According to the above control, the imitation personality generation device 2 (generative AI) executes a personality diagnosis in the form of free conversation for each of the user U and the imitation personality C, and the service management device 1 compares the personality diagnosis information of the user U and the personality diagnosis information of the imitation personality C, and judges the imitation accuracy of the imitation personality C from the personality similarity. Therefore, an imitation personality C that more accurately imitates the personality of the user U can be generated, and the effect of being able to reproduce the personality with high accuracy specialized for the user U is achieved.

[0087] Next, an example of improving the personality evaluation of an arbitrary user Un in the operation of the personality evaluation system S according to the embodiment will be described with reference to FIGS. 6 and 7.

[0088] As shown in FIG. 6, while user Un is in the middle of executing a chat with another user U (step E01: YES), the service management device 1 (the imitation personality execution unit 17) activates all the imitation personalities C1 to C(n - 1) stored in the imitation personality DB 20, and causes all the imitation personalities C1 to C(n - 1) to monitor the user behavior information 7n of user Un (step E02). Then, the service management device 1 (the imitation personality execution unit 17) causes all the imitation personalities C1 to C(n - 1) to evaluate whether each of a large number of behavior items recalled from these user behavior information 7n is "virtue", "landmine", or "no problem", and accumulates these evaluations for each of the imitation personalities C1 to C(n - 1) (step E03).

[0089] As described above, the thinking and judgment criteria of each of the imitation personalities C1 to C(n - 1) are made to be individually different like those of humans. For this reason, whether to give a positive evaluation, a negative evaluation, or a zero evaluation to the behavior items of each user U is divided by each imitation personality C (different judgments are made by each of the imitation personalities C1 to C(n - 1)).

[0090] When the evaluation by each of the imitation personalities C1 to C(n - 1) for all the recalled behavior items is completed (step E04: YES), the value waveform generation unit 24 of the service management device 1 calculates a value waveform (which can also be said to be a value graph), as a characteristic waveform showing the cumulative results of virtue and landmine for each of the imitation personalities C1 to C(n - 1) with respect to user Un (step E05). The value waveform in the embodiment is represented by a combination of a virtue waveform G+(Un) and a landmine waveform G-(Un). FIG. 7 shows an example of the value waveforms G+(Un), G-(Un) of user Un evaluated by the imitation personalities C1 to C(n - 1). The plus side of the vertical axis in the value waveforms G+(Un), G-(Un) is the cumulative value of virtue, and the minus side is the cumulative value of landmine.

[0091] When the value waveform G+(U) and G-(U) are created in this way, it is possible to project on the value waveforms G+(U) and G-(U) how the user U has been evaluated by each of the imitation personalities C1 to Cn-1, and it is possible to grasp the surface and inner personalities of each user U with higher accuracy. The surface and inner personalities of each user U can be visualized with high reliability. Users U with similar shapes of the value waveforms G+(U) and G-(U) can be evaluated as having similar (close) values. Users U with similar shapes of the value waveforms G+(U) and G-(U) may be expressed as having a high synchronization rate (which can also be called a similarity rate). Users U with completely different shapes of the value waveforms G+(U) and G-(U) can be evaluated as having different values. Users U with completely different shapes of the value waveforms G+(U) and G-(U) may be expressed as having a low synchronization rate.

[0092] Note that what shows the cumulative results of virtue and landmines judged for each imitation personality C with respect to the user U is not limited to the aforementioned value waveforms G+(U) and G-(U). Instead of the value waveform, for example, a value color chart may be adopted that expresses the cumulative results of virtue and landmines judged for each imitation personality C with respect to the user U in a gradation using each color such as "red" that associates victory, passion, love, etc., "yellow" that associates cheerfulness, youth, joy, etc., "green" that associates nature, harmony, peace, etc., and "blue" that associates silence, intellect, coldness, etc.

[0093] Next, an example of an approval process for determining whether to allow an arbitrary user Un to participate in a specific community SC to which users U1 to U(n-1) belong among the operations of the personality evaluation system S according to the embodiment will be described with reference to FIG. 8. Note that as the specific community SC, various communities such as a corporate community, a fan community, an online salon, a knowledge community, a hobby community, or a regional community may be adopted.

[0094] Here, as the value waveform G+(Un), G-(Un) of user Un, the graphs shown in FIG. 7 are obtained in advance. Also, in order to represent what kind of set of users U1 to U(n-1) with what kind of values the specific community SC is, the value waveforms G+(SC), G-(SC) of the specific community SC obtained by synthesizing the value waveforms G+(U), G-(U) of each user U are used. The value waveforms G+(SC), G-(SC) correspond to the characteristic waveforms of the specific community SC.

[0095] The value waveforms G+(SC), G-(SC) of the specific community SC are calculated by the service management device 1 (value waveform generation unit 24) based on the value waveforms G+(U), G-(U) of each user U and stored in the storage unit 13. As an example of calculating the value waveforms G+(SC), G-(SC) of the specific community SC, for example, for the cumulative value for each imitation personality C among the virtue waveforms G+(U) of each user U, the most frequent value, the median value, or the average value is taken to calculate the virtue waveform G+(SC), and for the cumulative value for each imitation personality C among the mine waveforms G-(U) of each user U, the most frequent value, the median value, or the average value is taken to calculate the mine waveform G-(SC), then the value waveforms G+(SC), G-(SC) of the specific community SC can be obtained.

[0096] As shown in FIG. 8, when user Un starts the chat application 5 on his / her user terminal 3 and applies to join the specific community SC (step P01: YES), the service management device 1 (waveform similarity calculation unit 16) reads out the value waveforms G+(Un), G-(Un) of user Un and the value waveforms G+(SC), G-(SC) of the specific community SC from the storage unit 13 (step P02). Next, the service management device 1 (waveform similarity calculation unit 16) compares the value waveforms G+(Un), G-(Un) of user Un with the value waveforms G+(SC), G-(SC) of the specific community SC and calculates the waveform similarity of user Un with respect to the specific community SC (step P03).

[0097] Next, if the waveform similarity of user Un with respect to specific community SC exceeds a predetermined threshold (step P04: YES), the value waveform G+(Un), G-(Un) of user Un is considered to be similar to (highly synchronized with) the value waveforms G+(SC), G-(SC) of specific community SC, and it is evaluated that their values match. Therefore, the service management device 1 (waveform similarity calculation unit 16) issues a command to permit user Un to participate in specific community SC to user terminal 3 (chat application 5) of user Un (step P05), enabling it to chat with other users U.

[0098] If the waveform similarity of user Un with respect to specific community SC is below the predetermined threshold (step P04: NO), the value waveforms G+(Un), G-(Un) of user Un are considered to be different from (lowly synchronized with) the value waveforms G+(SC), G-(SC) of specific community SC, and it is evaluated that their values do not match. Therefore, the service management device 1 (waveform similarity calculation unit 16) issues a command to reject user Un's participation in specific community SC to user terminal 3 (chat application 5) of user Un (step P06), making user Un unable to participate in specific community SC. In step P06, an example of rejecting user Un's participation in a specific community is shown, but it is not limited to this. Service providers may set countermeasures (such as setting various restrictions in specific community SC) based on the evaluation that their values do not match.

[0099] According to the above control, for example, even when inviting user U whose values do not match and who is highly likely to cause trouble, user U can smoothly and without coercion reject participation in specific community SC. Therefore, troubles among users U can be surely prevented. Also, since excellent users U whose values match and who are considered to have good personalities will gather, it is easy to activate specific community SC, and it can develop into a safe and secure specific community SC. The reliability of specific community SC is also improved.

[0100] Note that a user Un who has once been rejected from participating in a specific community SC is not necessarily permanently prohibited from participating in the specific community SC. The value waveform G+(Un), G-(Un) of the user Un will naturally change according to the accumulation of actions after the rejection of participation. As a result, if the value waveforms G+(Un), G-(Un) of the user Un come to resemble the value waveforms G+(SC), G-(SC) of the specific community SC and are evaluated as having matching (similar, close) values, the user Un can participate in the specific community SC.

[0101] In the embodiment shown in FIG. 8, an example of the comparison (approval process) of the degree of similarity in the relationship between the user U and the specific community SC is shown. However, it is not limited to this, and the comparison of the degree of similarity in the relationship between the user U and another user U can be realized in the same manner as the embodiment shown in FIG. 8, and the comparison of the degree of similarity in the relationship between the specific community SC and another specific community SC can also be realized in the same manner as the embodiment shown in FIG. 8.

[0102] For example, in the comparison of the degree of similarity in the relationship between a specific community SC and another specific community SC, if their values match (the degree of similarity is high), it is possible to recommend the other specific community SC to each other for mutual cooperation when searching, for example. Conversely, if their values do not match (the degree of similarity is low), it is possible not to introduce the other specific community SC even when searching.

[0103] The configuration of each part in the present invention is not limited to the illustrated embodiment, and various modifications can be made without departing from the spirit of the present invention.

Explanation of Reference Numerals

[0104] C, C(n-1), Cn Imitative Personality G+(U), G+(Un) Virtue Waveform G-(U), G-(Un) Mine Waveform S Personality Evaluation System SC Specific Community U, U(n-1), Un User 1 Service management device 2 Generative AI 3, 3n User terminal 4 Communication network 5 Chat app 6, 6n User ID 7, 7n User behavior information 11 Communication unit 12 Information processing unit 13 Memory unit 14, 38 System bus 15 Personality similarity calculation unit 16 Waveform similarity calculation unit 17 Imitative personality execution unit 18 User information DB 19 User behavior information DB 20 Imitative personality DB 21 Imitative personality generation unit 22 Skill correction unit 23 Value correction unit 24 Value waveform generation unit 31 CPU 32 Memory 33 Storage 34 Communication IF 35 Input device 36 Display unit 37 Input / output IF

Claims

1. The present invention includes a service management device that provides and manages a conversation service for a user terminal, and an imitation personality generation device that generates a virtual imitation personality corresponding to a user who uses the user terminal, The mimic personality generation device includes a mimic personality generation unit that generates a mimic personality that reflects the personality of the user based on conversation information of the user, a skill correction unit that corrects the mimic personality in response to skills possessed by the user, and a value correction unit that corrects the mimic personality in response to a threshold value of the user's value obtained from the conversation information of the user, the imitation personality generation device executes a personality diagnosis in a free conversation format for each of the user and the imitation personality; the service management device compares the personality diagnostic information of the user with the personality diagnostic information of the imitated personality, and determines the imitation accuracy of the imitated personality based on the degree of similarity between the personality diagnostic information of the imitated personality and the imitated personality; Personality assessment system.

2. The service management device has a large number of the mimic personalities for each user, having the plurality of imitative personalities evaluate the speech and behavior of a given user obtained from conversation information of the given user, and setting the evaluation result as a characteristic waveform of the given user; 2. A personality evaluation system according to claim 1.

3. The service management device compares a characteristic waveform of the arbitrary user with a characteristic waveform of another arbitrary user, and determines whether the two users match based on the degree of similarity.

3. A personality evaluation system according to claim 2.

4. The service management device sets a collection of characteristic waveforms of a plurality of arbitrary users as a characteristic waveform of a specific community, comparing a characteristic waveform of the given user with a characteristic waveform of the specific community, and judging whether the given user matches the specific community based on the degree of similarity; 3. A personality evaluation system according to claim 2.

5. The service management device sets a collection of characteristic waveforms of a plurality of arbitrary users as a characteristic waveform of a specific community, comparing a characteristic waveform of the specific community with a characteristic waveform of another specific community, and judging whether the two specific communities match based on the degree of similarity; 3. A personality evaluation system according to claim 2.

Images