How to Converse

Abstract

To provide a personality formation method capable of easily forming the personality of a virtual person and of conversing with a virtual person, especially such as a dead person.SOLUTION: A personality formation method that a computer forms the personality of a virtual person comprises the steps of acquiring advance fundamental data for the virtual person, asking the chatbot using a large-scale language model to interview the virtual person based on the acquired advance fundamental data, and forming the personality model of the virtual person based on answers to the questions. A conversation method comprises the steps of generating an answer to a question based on this personality model and outputting the generated answer as a text or an audio.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a technique that is effective in forming the personality of a virtual character. [Background technology]

[0002] Conventionally, methods for creating a virtual personality using a computer have been known. For example, Patent Document 1 discloses a technique for storing the video and audio of a real person and outputting a virtual person using a predetermined video model and personality model. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2021-088618 Summary of the Invention [Problem to be solved by the invention]

[0004] However, in order to create a personality model of a virtual character using artificial intelligence, it is necessary to prepare a large amount of data that indicates the character's personality, so it has been practically difficult to create a virtual character's personality using learning data and have the virtual character perform functions like a chatbot.

[0005] An object of the present invention is to provide a personality formation method, a conversation method, a personality formation system, and a program that enable a person to easily create the personality of a virtual character and to converse with a virtual character, particularly a deceased person. [Means for solving the problem]

[0006] The present invention ,Ko On a computer By repeating questions and answers to create a virtual person and enable conversation with that virtual person. conversation of A conversation method comprising: acquiring personal information and attribute information about the virtual person as prior basic data for the virtual person; A chatbot utilizing a large-scale language model conducts an interview with the virtual person along with prior basic data about the virtual person. thing and The chatbot creates a virtual person based on the preliminary basic data. creating a personality model for the virtual person based on responses from a given user to an interview; Based on the personality model formed, From a real person generating an answer to the question; outputting the generated answer in text or voice; A conversation method is provided.

[0007] According to the present invention, prior basic data about a virtual person is used to ask a chatbot utilizing a large-scale language model questions about conducting an interview with the virtual person, and a personality model of the virtual person is created based on the answers to these questions. Therefore, by easily generating learning data for forming the personality of a virtual character, a personality model can be formed, and by using this personality model, it is possible to have conversations with virtual characters, especially dead people.

[0008] Although the present invention is in the category of a method, the same functions and effects specific to the category can be achieved even when it is applied to a system or a program.

[0009] Furthermore, according to the present invention, answers to questions are generated based on the created personality model and the generated answers are output in text or voice, thereby enabling conversation with virtual characters, particularly the dead. [Effects of the Invention]

[0010] According to the present invention, it is possible to easily create the personality of a virtual person and to converse with a virtual person, particularly a dead person. [Brief explanation of the drawings]

[0011] [Figure 1] FIG. 1 is a diagram illustrating an overview of a personality development system 1. [Figure 2] (2a) is a schematic diagram of the entire system, and (2b) is a diagram showing the tuning device 10 in isolation. [Figure 3] FIG. 2 is a flowchart showing the personality model formation process and the conversation process based on this personality model executed by the personality formation system 1. [Figure 4] FIG. 2 is a flowchart showing the personality model formation process executed by the personality formation system 1. [Figure 5] FIG. 10 is a diagram schematically illustrating an example of a screen when a personality model is formed. [Figure 6] FIG. 10 is a flowchart showing a conversation process based on a personality model executed by the personality development system 1. [Figure 7] FIG. 10 is a diagram schematically illustrating an example of a screen during a conversation. [Figure 8] FIG. 10 is a flowchart showing the personality model update process executed by the personality formation system 1. [Figure 9] FIG. 10 is a diagram schematically illustrating an example of an application screen. [Figure 10] FIG. 10 is a diagram schematically illustrating an example of an application screen. [Figure 11] FIG. 10 is a diagram schematically illustrating an example of an application screen. DETAILED DESCRIPTION OF THE INVENTION

[0012] Hereinafter, embodiments of the present invention (hereinafter referred to as "embodiments") will be described in detail with reference to the accompanying drawings. In the following drawings, the same elements are designated by the same numbers or symbols throughout the description of the embodiments.

[0013] Fig. 1 is a block diagram illustrating an outline of the instruction configuration of a personality development system 1. Fig. 2a is a block diagram illustrating an outline of the personality development system 1, and Fig. 2b is a diagram illustrating a tuning device 10. The components of the personality development system 1 will be described with reference to FIGS.

[0014] The personality development system 1 is a system for developing a personality for a virtual character, and includes at least a tuning device 10. In this embodiment, the personality development system 1 further includes, in addition to the tuning device 10, a personality model character terminal 3 used by the personality model character 2 and a real character terminal 5 used by the real character 4. The tuning device 10 is connected to each of the personality model character terminal 3 and the real character terminal 5 so as to be able to communicate data. Here, a virtual character is a living being with a personality, such as an imaginary person, and may be a deceased person who has already passed away, or may be a living person. It may also be a virtual character from an anime or the like, or an object with a fantasy personality.

[0015] The personality model person terminal 3 is a terminal device used by the personality model person 2, such as a mobile phone, a smartphone, a tablet terminal, a personal computer, or a laptop computer. The real person terminal 5 is a terminal device used by the real person 4, such as a mobile phone, a smartphone, a tablet terminal, a personal computer, or a laptop computer.

[0016] The tuning device 10 includes, for example, a processor, a communication unit, an output unit that outputs signals, etc., an acquisition unit that acquires prior basic data for the virtual person, a questioning unit that asks a chatbot utilizing a large-scale language model to interview the virtual person based on the acquired prior basic data, and a creation unit that creates a personality model of the virtual person based on the answers to the questions, and may be realized by a single computer or by multiple computers such as a cloud computer. Note that the cloud computer in this specification may refer to either a computer that uses any computer in a scalable manner to perform a specific function, or a computer that includes multiple functional modules to realize a system and uses the functions in any combination. The tuning device 10 includes a storage unit such as a hard disk, semiconductor memory, recording medium, or memory card for storing data. The tuning device 10 includes, as a processor, various devices that execute various processes. The tuning device 10 is connected to the personality model person terminal 3 and the real person terminal 5 via a network 8.

[0017] [Device configuration] As shown in FIG. 2, the personality development system 1 is a system for developing the personality of a virtual character, and a tuning device 10 is connected to the system so as to be able to communicate data via a network 8 such as a public line network. The components of the personality development system 1 are merely examples, and the number, types, and functions of terminals and devices not shown can be changed as appropriate.

[0018] The personality model person terminal 3 and the real person terminal 5 are equipped with a CPU (Central Processing Unit), GPU (Graphics Processing Unit), RAM (Random Access Memory), ROM (Read Only Memory), etc. as terminal control units, and are equipped with devices, etc. as communication units to enable communication with other terminals and devices, etc.

[0019] The tuning device 10 includes a CPU, GPU, RAM, ROM, etc. as a processor, and includes a device, etc. as a communication unit, for enabling communication with other terminals, devices, etc.

[0020] The processing steps performed by the personality formation system 1 when forming a personality for a virtual character will be outlined with reference to FIGS.

[0021] The tuning device 10 acquires preliminary basic data for a virtual person (step S1). The prior basic data is data about a person who is the basis for the virtual character. The prior basic data is, for example, personal information (information that can identify an individual, such as name, date of birth, and address) and attribute information (information about the nature and characteristics of an individual, such as age, sex, place of residence, family structure, occupation, educational background, and income). The tuning device 10 acquires this prior basic data from the personality model person terminal 3.

[0022] In addition, if the information about the costumed person can be obtained in advance as open information from the Internet or the like, the tuning device 10 may obtain this open information in advance from the Internet or the like, and obtain only the information that cannot be obtained as open information from the personality model person terminal 3.

[0023] Based on the acquired prior basic data, the tuning device 10 asks the chatbot utilizing the large-scale language model to conduct an interview with a virtual person (step S2). Based on the acquired prior basic data, the tuning device 10 asks a chatbot utilizing a large-scale language model (interviewer model in FIG. 1) a question about conducting an interview with a virtual person. This question is generated based on the prior basic data using the large-scale language model. This question may include not only text information but also audio, images, videos, etc. The tuning device 10 outputs text or voice based on the generated question to the personality model person terminal 3.

[0024] The tuning device 10 creates a personality model of the virtual person based on the answers to the questions (step S3). The tuning device 10 receives an answer to the output question from the personality model person terminal 3. This answer may be not only text information but also audio, image, video, or the like. The tuning device 10 adds the generated questions and the answers to these questions to a data set, performs fine tuning, and forms a personality model of the virtual person.

[0025] In the processing of steps S1 to S3, the target person for the virtual person is a person who has already passed away (deceased) or a person who is currently alive (living person). When a deceased person is the target, it is sufficient to use the prior basic data and answers to questions input by a close relative of the deceased into the personality model person terminal 3 using text, voice, etc. When a living person is the target, it is sufficient to use the prior basic data and answers to questions input by the living person or a close relative of the living person into the personality model person terminal 3 using text, voice, etc. Using a personality model of a living person after their death results in a personality model of the deceased.

[0026] The tuning device 10 receives a question from the real person 4 (step S4). The tuning device 10 accepts a question from a real person 4 via a real person terminal 5.

[0027] Tuning device 10 generates an answer to the question based on the formed personality model (step S5). The tuning device 10 uses the personality model that has been created to generate answers to received questions.

[0028] The tuning device 10 outputs the generated answer in text or voice (step S6). The tuning device 10 outputs text or voice based on the generated answer to the real person terminal 5.

[0029] The tuning device 10 stores the received questions and generated answers in chronological order as a conversation history. When the conversation history satisfies a predetermined condition (the number of tokens in the conversation history exceeds a predetermined value, every predetermined time elapses, etc.), the tuning device 10 adds the conversation history to a dataset, performs fine tuning, and updates the personality model.

[0030] The processing steps that the personality formation system 1 performs to form a personality of a virtual character in steps S1 to S3 will be outlined with reference to FIGS.

[0031] The tuning device 10 acquires, in the acquisition unit, prior basic data for the virtual person (step S10). The virtual person is a fictitious person based on the personality model person 2. For example, if the virtual person is a deceased person, the personality model person 2 may be a person who knows the deceased in detail, such as a close relative of the deceased or a person who was close to the deceased (a person who knows the hobbies, tastes, ideas, beliefs, words, actions, thoughts, etc.). If the virtual person is a living person, the personality model person 2 may be the living person himself or a person who knows the living person in detail, such as a close relative of the living person or a person who was close to the living person (a person who knows the hobbies, tastes, ideas, beliefs, words, actions, thoughts, etc.). In this embodiment, the virtual person will be described in the processing described below as particularly referring to a deceased person. This is because, even if the personality model person 2 was a living person when the preliminary basic data was acquired, he or she will already be deceased by the time the processing described below is actually performed. The prior basic data is data related to a virtual person. Examples of the prior basic data include personal information (such as name, date of birth, and address) that can identify an individual, and attribute information (such as age, sex, place of residence, family structure, work history, educational background, hobbies, preferences, thoughts, beliefs, behavior, and thoughts) that is information related to the individual's attributes. The personality model person terminal 3 accepts input of prior basic data for the virtual character. The personality model person 2 inputs the prior basic data of the virtual character using a virtual keyboard, a text file based on a format, or the like. The personality model person terminal 3 transmits the accepted prior basic data to the tuning device 10. The acquisition unit receives this preliminary basic data and acquires the preliminary basic data for the virtual person.

[0032] The tuning device 10 may acquire this preliminary basic data from a social networking service (SNS) or a predetermined website (such as a site provided by mass media or a curation site). In this case, the tuning device 10 acquires the personal information of the virtual character and the ID of the virtual character on the SNS by a predetermined method (such as from the personality model person terminal 3), and acquires the content that the virtual character has posted to the SNS in the past (including not only posts by the virtual character himself, but also responses from other characters, such as replies and quotes to these posts), and content related to the virtual character on a predetermined website (not only content directly related to the virtual character (such as articles related to the virtual character), but also content related to the personal information of the virtual character (such as content around the residence (events, weather, incidents, etc.) and content of the affiliation included in the work history (such as new business ventures, business conditions, etc.)).

[0033] The tuning device 10 can also be configured to output the acquired prior basic data to the personality model person terminal 3 and have the personality model person 2 confirm the accuracy of the content. In this case, the tuning device 10 may, for example, perform an online search for each piece of information included in the acquired prior basic data and output the search results, such as a website, to the personality model person terminal 3. Specifically, if one of the work histories in the prior basic data is "X Corporation," the tuning device 10 performs an online search for "X Corporation" and outputs the search results, such as "X Corporation's website," along with predetermined text. As an example, the tuning device 10 outputs text such as "Is X Corporation the company you belong to? http: / / www.xxx.com (this may be a URL (Uniform Resource Locator) or may have the website embedded)" and the website. The tuning device 10 acquires a response to the output result (whether it is correct or incorrect, etc.) from the personality model person terminal 3. If the response to the output result is correct, the tuning device 10 treats this information as prior basic data. On the other hand, if the response to the output result is incorrect, the tuning device 10 may exclude this information from the prior basic data, or if the personality model person terminal 3 has sent correct information in addition to the incorrect information, the tuning device 10 may update the confirmed information to the correct information and treat the updated information as prior basic data.

[0034] The tuning device 10, in the questioning section, asks the chatbot utilizing the large-scale language model to conduct an interview with a virtual person based on the acquired prior basic data (step S11). The questioning unit includes a chatbot and has the function of asking predetermined questions to the chatbot. The questioning unit asks the chatbot, which utilizes a large-scale language model (the interviewer model in Figure 1), a question about conducting an interview with a virtual person, along with the acquired prior basic data. Here, the questioning unit inputs to the chatbot information about conducting an interview with a virtual person, the acquired prior basic data, how to ask the question, etc. The content that the questioning unit inputs to the chatbot is not limited to these examples. In this embodiment, the tuning device 10 is described as holding a large-scale language model, but this large-scale language model may be held by another computer having a server function, etc. In this case, the tuning device 10 outputs the acquired prior basic data to this computer, which then executes a function corresponding to the interrogator and outputs the execution result to the tuning device 10.

[0035] The tuning device 10 outputs the question generated by the chatbot (step S12). The chatbot generates questions based on the input. The tuning device 10 outputs the question generated by this chatbot to the personality model person terminal 3. Here, the tuning device 10 transmits this question in either text format or voice format, or in both formats. The personality model person terminal 3 receives this question and displays it in text format on its own display unit (monitor, touch panel, etc.), or outputs it in audio format through its audio unit (microphone, etc.).

[0036] The tuning device 10 receives an answer to the question (step S13). The personality model person terminal 3 receives an input of an answer to a question from the personality model person 2. The personality model person terminal 3 receives this answer in either text format or voice format, or both. For example, it receives text or voice input into a predetermined input form. The personality model person terminal 3 transmits the received answer to the tuning device 10. Tuning device 10 receives this response and accepts the answer to the question.

[0037] An example of a screen displayed during the processing of steps S11 to S13 will be described with reference to Fig. 5. The figure is a diagram schematically showing an example of a screen displayed during the processing of steps S11 to S13. In the figure, an input screen 20 is shown. This input screen 20 is an example of the personality model person terminal 3 displaying questions and answers in text format. The input screen 20 has an input field 21 that accepts input of answers from the personality model person 2, and a first field 22, a second field 23, a third field 24, and a fourth field 25 that display questions, answers, etc. The contents of the first field 22 and the third field 24 are based on the content entered into the input field 21. The contents of the second field 23 and the fourth field 25 are based on questions generated by the chatbot. Note that the content of the first field 22 may be based on the content entered by the question section. In the first column 22, the virtual person to be interviewed (Please interview the following person. Please fill in any missing information by asking questions), preliminary basic data (name, date of birth, birthplace, educational background, place of employment), and how to ask questions (example questions) are shown. The personality model person terminal 3 accepts these inputs in the input column 21, displays the accepted content in the first column 22, and transmits it to the tuning device 10. Based on this content and the preliminary basic data, the tuning device 10 asks the chatbot utilizing the large-scale language model to interview the personality model person 2. The tuning device 10 outputs the question generated by the chatbot to the personality model person terminal 3, and the personality model person terminal 3 displays the question generated by the chatbot in the second column 23. The personality model person terminal 3 accepts input of an answer to the question shown in the second column 23 in the input column 21, displays the accepted answer in the third column 24, and transmits it to the tuning device 10. Based on this content, the tuning device 10 asks the chatbot utilizing the large-scale language model further questions about the answer or questions about the lack of prior basic data. The tuning device 10 outputs the questions generated by the chatbot and the details of the missing preliminary basic data to the personality model person terminal 3, and the personality model person terminal 3 displays the generated questions and the details of the missing preliminary basic data in the fourth column 25. The tuning device 10 and the personality model person terminal 3 repeat the exchange of questions and answers until the amount of data required for generating a personality model is met.

[0038] The tuning device 10 determines whether or not a predetermined condition is met (step S14). The predetermined condition is, for example, whether or not the amount of data is necessary for forming a personality model. The tuning device 10 determines, for example, whether or not a preset amount of prior basic data and an answer have been received. If the tuning device 10 determines that the predetermined conditions are not met (step S14 NO), for example, if it determines that the received preliminary basic data and answers do not amount to a sufficient amount of data, the tuning device 10, in the questioning section, asks the chatbot utilizing the large-scale language model to conduct a further interview with the virtual person based on the received answer (step S15). The questioning unit asks the chatbot, which utilizes a large-scale language model, questions about conducting further interviews based on the received answers, and inputs the received answers to the chatbot. Thereafter, the tuning device 10 executes the processes from step S12 onwards again. That is, if the amount of data in the preliminary basic data or answers is insufficient, the tuning device 10 outputs a question newly generated by the chatbot based on the received answers, etc., to the personality model person terminal 3, and receives an answer to this question from the personality model person terminal 3. The tuning device 10 and the personality model person terminal 3 repeat the exchange of questions and answers until the preset data amount is met.

[0039] If the tuning device 10 determines that a predetermined condition is met (YES in step S14), for example, if it determines that the received preliminary basic data and answers have reached a sufficient amount of data, the tuning device 10 forms a personality model of the virtual person in the formation unit based on the answers to the questions (step S16). The formation unit uses the preliminary basic data and the received answers as a data set, performs fine tuning, and forms a personality model of personality model person 2. Tuning device 10 stores the created personality model in its own storage unit.

[0040] The above is an outline of the processing steps for forming the personality of a virtual character in steps S1 to S3.

[0041] When forming a personality model, the tuning device 10 may form the personality model by age group for the same personality model person 2. For example, the tuning device 10 may form personality models of virtual characters by age group, such as 20s, 30s, 40s, 50s, and 60s. In this case, the personality model person 2 may be a living person of that age group, or a close relative who knew a living or deceased person of that age group. The tuning device 10 executes the above-described process for each generation and forms a personality model for each generation. That is, the tuning device 10 executes the above-described process at predetermined intervals (every year, every five years, or every ten years). When the personality model person 2 reaches a different generation, the tuning device 10 forms a new personality model and stores the personality model for each generation. When having a conversation using personality models for different age groups, the real person 4 can converse with virtual people of different age groups. As a result, the real person 4 can more naturally discuss worries, questions, etc. with virtual people of a similar age to his or her own.

[0042] In the processing of steps S10 to S16 described above, the target person for the virtual person is a person who has already passed away (deceased person) or a person who is currently alive (living person). When a deceased person is the target, the prior basic data, answers, etc. may be provided by a close relative of the deceased person. When a living person is the target, the prior basic data, answers, etc. may be provided by the living person themselves or a close relative of the living person. If the personality model formed is a deceased person, when a conversation is conducted using the personality model, the conversation will be conducted virtually with the deceased. If the personality model formed is a living person, when a conversation is conducted using the personality model, if the living person has already passed away for some reason, the conversation will essentially be synonymous with the personality model of the deceased, and the conversation will be conducted virtually with the deceased. Furthermore, even if the living person is not dead, the conversation will be with a personality from a past time. Furthermore, if the living person is not dead, the third party will be conducting a virtual conversation rather than a conversation with a real living person.

[0043] An outline of the processing steps when the personality formation system 1 conducts a conversation using the formed personality model in steps S4 to S6 will be described with reference to FIGS.

[0044] The tuning device 10 receives a question from the real person 4 (step S20). The real person terminal 5 accepts a question input from the real person 4. The real person terminal 5 accepts this question in either text format or voice format, or both. For example, it accepts text or voice input into a predetermined input form. The real person terminal 5 transmits the accepted question to the tuning device 10. The tuning device 10 receives this question and accepts a question from the real person 4 .

[0045] Tuning device 10 generates an answer to the question based on the formed personality model and the received question (step S21). Tuning device 10 uses this personality model to generate an answer to the received question. Tuning device 10 performs text recognition and speech recognition on the content of the question and analyzes the content of the question. Tuning device 10 references predetermined keywords and the like included in the analysis results to the personality model to generate an answer to the question. Here, the answer generated by tuning device 10 is one that leaves a positive impression (gratitude, praise, etc.) other than those that give a negative impression to real person 4 (complaints, slander, criticism, etc.). The tuning device 10 may generate only one answer for one question, or may generate different answers for one question. That is, the tuning device 10 may always generate the same answer for a certain question, or may generate a different answer each time a question is received. The different answers may be, for example, answers that conform to the personality (way of thinking, tone of voice, etc.) of the virtual character in the personality model, or answers that are contrary to the personality or different from the personality.

[0046] The tuning device 10 outputs the generated answer (step S22). The tuning device 10 outputs the generated answer to the real person terminal 5. Here, the tuning device 10 transmits this answer in either text format or audio format, or both. The real person terminal 5 receives this response and displays it in text format on its own display unit (monitor, touch panel, etc.), or outputs it in audio format through its audio unit (microphone, etc.).

[0047] The tuning device 10 repeats the processing of steps S20 to S22 described above until the real person 4 finishes asking questions, thereby causing a conversation to take place between the virtual person (dead person) and the real person (living person) 4.

[0048] An example of a screen displayed during the processing of steps S20 to S22 will be described with reference to Fig. 7. This figure is a diagram schematically showing an example of a screen displayed during the processing of steps S20 to S22. In this figure, a conversation screen 30 is shown. This conversation screen 30 is an example of the real person terminal 5 displaying questions and answers in text format. The conversation screen 30 has an input field 31 that accepts input of a question from a real person 4, and a first field 32, a second field 33, a third field 34, and a fourth field 35 that display the question and answer. The contents of the first field 32 and the third field 34 are based on the content entered into the input field 31. The contents of the second field 33 and the fourth field 35 are based on the answer generated based on the personality model and the question. The real person terminal 5 accepts a question input into the input field 31, displays the accepted question in the first field 32, and transmits the question to the tuning device 10. The tuning device 10 generates an answer based on the personality model and its contents. Tuning device 10 outputs the generated answer to real person terminal 5, and real person terminal 5 displays the generated answer in second column 33. The real person terminal 5 accepts input of a question in response to the answer shown in the second column 33 into the input column 31, displays the accepted question in the third column 34, and transmits the question to the tuning device 10. The tuning device 10 generates an answer based on the personality model and its contents. Tuning device 10 outputs the generated answer to real person terminal 5, and real person terminal 5 displays the generated answer in fourth column 35. The tuning device 10 and the real person terminal 5 repeat the exchange of questions and answers until the real person 4 finishes, thereby allowing a conversation to take place between the virtual person (dead person) and the real person (living person) 4.

[0049] The above is an outline of the processing steps when a conversation is held using the formed personality model in steps S4 to S6.

[0050] The outline of the processing steps when the personality formation system 1 updates the formed personality model will be described with reference to FIG.

[0051] The tuning device 10 stores the conversation history (step S30). The conversation history is a history of combinations of questions received in the processing of steps S20 to S22 and answers generated. Here, the conversation history may indicate a set of one question and one answer generated for this one question, or may indicate a series of questions from when the real person 4 started asking questions to when it finished asking questions, and sets of answers generated for this series of questions. Tuning device 10 stores the conversation history in chronological order.

[0052] The tuning device 10 determines whether the stored conversation history satisfies a predetermined condition (step S31). The predetermined condition is, for example, each time the number of tokens in the conversation history exceeds a certain value or a certain time period has elapsed. If the tuning device 10 determines that this conversation history does not satisfy a predetermined condition (step S31 NO), the process ends.

[0053] If the tuning device 10 determines that this conversation history satisfies a predetermined condition (YES in step S31), it adds the conversation content to the data set (step S32). The tuning device 10 adds the conversation content (a set of questions and answers to these questions) indicating the content of this conversation history to the data set used when forming the personality model.

[0054] The tuning device 10 updates the personality model based on the data set to which the conversation content has been added (step S33). The tuning device 10 fine-tunes the personality model based on the data set to which the conversation content has been added, and updates the personality model. Tuning device 10 updates the stored personality model by overwriting it with the updated personality model.

[0055] The above is an outline of the processing steps when updating a personality model that has been formed.

[0056] After executing these processes, when the personality development system 1 executes the processes of steps S20 to S22 again, it executes each process using the updated personality model. Furthermore, the processing of steps S30 to S33 is performed periodically, making it possible to have a conversation that takes into account the questions and answers asked by the real person 4. As a result, adding new information to the personality model not only makes the conversation more natural, but also makes it easier to keep the answers generated by the personality model up to date. For example, the personality development system 1 can generate a response based on the content of a previous conversation and carry out a conversation based on this response.

[0057] In addition to the conversation content, the tuning device 10 may also acquire content that the virtual person has posted to an SNS in the past and content related to the virtual person on a specified website from an SNS or a specified website, similar to the prior basic data, and fine-tune the personality model based on the acquired content to update the personality model. For example, by obtaining the latest weather information from a specified website and adding this weather information to the personality model, it is possible to make conversations more natural, even when talking about world events. Furthermore, in the processing step of having a conversation using the personality model formed as described above, the tuning device 10 can also be configured to not accept questions from the real person 4, but to generate answers based on prior basic data (anniversaries such as birthdays), the contents of the conversation history (plans included in the conversation, etc.), and content obtained from SNS or a specified website (weather, incidents, accidents, etc.). For example, when acquiring content related to a major event (anniversary, disaster, incident, etc.), a response based on this event (such as "It rained heavily yesterday, are you okay?") may be generated and output. The real person terminal 5 notifies the real person 4 of the output response by push notification or the like, and the real person 4 then begins a conversation with the virtual character. In this way, the tuning device 10 can be configured to make spontaneous utterances without relying on input from the real person 4. This can reduce the real person 4's sense of loneliness, loss, etc.

[0058] Examples of screens of an application that actually uses the present invention will be described with reference to Figures 9, 10, and 11. Each figure is a diagram that schematically shows an example of a UI (User Interface) of the application displayed on the real person terminal 5. 9 is a diagram showing a typical example of a home screen of this application, in which a home screen 40 is shown. The home screen 40 has a first column 41 that displays data about the deceased (date of death, whether or not they can attend a memorial service, etc.). The real-person terminal 5 accepts input into the first column 41 and transitions to a screen corresponding to the data. FIG. 10 is a diagram showing a typical example of a conversation screen of this application. In the diagram, a conversation screen 50 is shown. The conversation screen 50 has a first column 51 that displays data related to the conversation with the deceased. The real person terminal 5 accepts input into the first column 51, transitions to the screen shown in Figure 7, and realizes a conversation between the real person 4 and the deceased. 11 is a diagram showing a typical example of a request screen for this application, in which a request screen 60 is shown. The request screen 60 has a first column 61 that displays data (such as name and facial photograph) about the virtual person on which the created personality model is based. The real person terminal 5 accepts input into the first column 61, and transitions to a screen for setting the virtual person in this first column 61 as a relative of the real person, etc.

[0059] The above is an overview of Personality Formation System 1. According to the personality formation system 1, it is possible to easily form the personality of a virtual person and to have a conversation with a virtual person, particularly a dead person.

[0060] The above-described means and functions are realized by a computer (including a CPU, an information processing device, and various terminals) reading and executing a predetermined program. The program may be provided, for example, from a computer via a network (Software as a Service (SaaS)) or as a cloud service. The program may also be provided in a form recorded on a computer-readable recording medium. In this case, the computer reads the program from the recording medium, transfers it to an internal or external recording device, records it, and executes it. The program may also be pre-recorded on a recording device (recording medium) and provided to the computer from the recording device via a communication line.

[0061] Although the embodiments of the present invention have been described above, the present invention is not limited to the above-described embodiments. Furthermore, the effects described in the embodiments of the present invention are merely a list of the most preferable effects resulting from the present invention, and the effects of the present invention are not limited to those described in the embodiments of the present invention.

[0062] A first aspect disclosed in this embodiment is a personality development method in which a computer develops a personality of a virtual character, acquiring prior basic data for the virtual person; A step of asking a chatbot utilizing a large-scale language model to conduct an interview with the virtual person based on the acquired prior basic data; creating a personality model for the virtual person based on answers to the questions; The present invention provides a personality development method comprising the steps of:

[0063] In a second aspect disclosed in this embodiment, the virtual person is a deceased person. A personality development method according to a first aspect is provided.

[0064] A third aspect disclosed in this embodiment is a method for generating a personality model based on the personality model formed in the step of forming the personality model according to the first aspect, generating an answer to the question; outputting the generated answer in text or voice; A conversation method is provided.

[0065] A fourth aspect disclosed in this embodiment is a personality development system for developing a personality of a virtual person, an acquisition unit that acquires prior basic data for the virtual person; a questioning unit that asks a chatbot utilizing a large-scale language model to conduct an interview with the virtual person based on the acquired prior basic data; a creation unit that creates a personality model of the virtual person based on the answers to the questions; The present invention provides a personality development system comprising:

[0066] A fifth aspect disclosed in this embodiment is a computer that forms a personality of a virtual person, acquiring prior basic data for the virtual person; a step of asking a chatbot utilizing a large-scale language model to conduct an interview with the virtual person based on the acquired prior basic data; creating a personality model for the virtual person based on answers to the questions; A computer-readable program for executing the above is provided. [Explanation of symbols]

[0067] 1 Personality Development System 2. Personality model 3. Personality Model Persona Terminal 4 Real people 5 Real-life terminal 8 Network 10 Tuning Device 20 Input screen 21 Input field 22 Column 1 23 Column 2 24 Column 3 25 Column 4 30 Conversation screen 31 Input field 32 Column 1 33 Column 2 34 Column 3 35 Column 4 40 Home Screen 41 Column 1 50 Conversation screen 51 Column 1 60 Request Screen 61 Column 1

Claims

1. A conversation method for forming a virtual person in a computer and having a conversation with the virtual person by repeating questions and answers, comprising: acquiring personal information and attribute information about the virtual person as prior basic data for the virtual person; A step of inputting, into a chatbot utilizing a large-scale language model, information about conducting an interview with the virtual person together with prior basic data about the virtual person; forming a personality model of the virtual person based on answers from a predetermined user to an interview with the virtual person formed by the chatbot based on the prior basic data; generating answers to questions from real people based on the personality model formed; outputting the generated answer in text or voice; A conversation method comprising:

2. 2. The conversation method according to claim 1, wherein the virtual person is a deceased person, and the response from the predetermined user is a response from a close relative of the deceased virtual person.

3. In the acquiring step, the preliminary basic data is acquired from a social networking site (SNS) or a website. The conversation method according to claim 1.

4. In the step of asking, the question is asked about the lack of the acquired preliminary basic data. The conversation method according to claim 1.

5. In the step of forming a personality model of the virtual person, a personality model of the virtual person is formed by age group. The conversation method according to claim 1.

6. In the generating step, not only answers that conform to the personality of the virtual person in the personality model but also answers that are contrary to or different from the personality are generated. The conversation method according to claim 1.

Images