Information processing program, information processing method, and information processing device

Abstract

An information processing device (100) generates first information (111) explaining content related to persona information (140) among items of question information (150). The information processing device (100) generates second information (121) explaining the meaning of the question information (150). On the basis of the first information (111) and the second information (121), the information processing device (100) generates third information (131) explaining an answer to a question corresponding to information (141) regarding each of one or more items. The information processing device (100) generates answer information (160) with respect to the question information (150) on the basis of the third information (131).

Description

Information Processing Program, Information Processing Method, and Information Processing Apparatus

[0001] The present invention relates to an information processing program, an information processing method, and an information processing apparatus.

[0002] Conventionally, in order to imitate the response of a specific person to a question, there is a technology of Character-LLM (Large Language Model) that has learned information such as a profile or life story of a specific person. An LLM is an AI (Artificial Intelligence) model related to natural language processing.

[0003] As prior art, for example, there is one that adds user utterances to a prompt including an instruction sentence related to an explanation of a chatbot character, encodes the prompt, and inputs it to a language model. Also, for example, there is a technology of acquiring a first context corresponding to utterance data from a set of context candidates and generating one or more dialogue sets including the first context and the utterance data. Also, for example, there is a technology of generating a response based on a personality comparison between a first character and a user. Also, for example, there is a technology of converting a user's voice input into text and analyzing the emotion expressed by the user.

[0004] Japanese Patent Application Laid-Open No. 2022-180282, Japanese Patent Application Laid-Open No. 2023-073220, US Patent Application Publication No. 2020 / 0137001, US Patent Application Publication No. 2024 / 0169974

[0005] However, in the prior art, it may be difficult to accurately imitate the response of a specific person to a question. For example, it is difficult to accurately imitate the response of a specific person to a question that appropriately reflects information such as a profile or life story of a specific person.

[0006] In one aspect, the present invention aims to accurately imitate the response of a specific person to a question.

[0007] According to one embodiment, an information processing program, information processing method, and information processing device are proposed that acquire question information relating to a question, acquire persona information including information for each of one or more items relating to the characteristics of a target, generate first information that explains the content related to the acquired persona information from the acquired question information, generate second information that explains the meaning of the acquired question information, generate third information that explains the response to the question according to the information for each of the one or more items based on the generated first information and the generated second information, and generate response information to the question information based on the generated third information.

[0008] According to one embodiment, it becomes possible to accurately mimic the response of a specific person to a question.

[0009] Figure 1 is an explanatory diagram showing one embodiment of the information processing method according to the embodiment. Figure 2 is an explanatory diagram showing an example of the information processing system 200. Figure 3 is a block diagram showing an example of the hardware configuration of the control device 201. Figure 4 is a block diagram showing an example of the functional configuration of the information processing device 100. Figure 5 is an explanatory diagram showing an example of persona information 500. Figure 6 is an explanatory diagram (part 1) showing an example of the processing in step 1. Figure 7 is an explanatory diagram (part 2) showing an example of the processing in step 1. Figure 8 is an explanatory diagram (part 1) showing an example of the processing in step 2. Figure 9 is an explanatory diagram (part 2) showing an example of the processing in step 2. Figure 10 is an explanatory diagram (part 3) showing an example of the processing in step 2. Figure 11 is an explanatory diagram (part 4) showing an example of the processing in step 2. Figure 12 is a flowchart showing an example of the overall processing procedure. Figure 13 is a flowchart showing an example of the generation processing procedure.

[0010] Embodiments of the information processing program, information processing method, and information processing apparatus according to the present invention will be described in detail below with reference to the drawings.

[0011] (An Embodiment of the Information Processing Method According to the Embodiment) Figure 1 is an explanatory diagram showing an embodiment of the information processing method according to the embodiment. The information processing device 100 is a computer that uses a language model to mimic the response of a specific person to a question. The information processing device 100 is, for example, a server or a PC (Personal Computer).

[0012] A language model is, for example, an LLM (Language Language Model). An LLM has the function of outputting text as a response to input text. An LLM is generated, for example, by learning from multiple texts. An LLM is, for example, a GPT (Generative Pretrained Transformer). GPT is a registered trademark.

[0013] For example, conventionally, there is a Character-LLM (Character-Likely Learning Model) technique that learns persona information about a specific person in order to mimic the responses of that person to a question. Learning is, for example, fine-tuning based on persona information. Persona information includes, for example, information about the attributes of a specific person. Specifically, persona information includes information such as a profile or life story about a specific person.

[0014] Furthermore, there is an LLM Roleplay technique that, for example, includes text in the initial prompt to the LLM requesting it to mimic a specific person, thereby controlling the LLM to mimic the responses of that specific person to questions in subsequent interactions. The initial prompt may, for example, include persona information.

[0015] However, regardless of the technology used, it can be difficult to accurately mimic a specific person's response to a question. For example, it can be difficult to accurately mimic a specific person's response to a question that properly reflects persona information. Specifically, when mimicking a specific person's response to a question, persona information is not taken into consideration, making it impossible to accurately mimic that person's response.

[0016] Therefore, this embodiment describes an information processing method that can accurately mimic the response of a specific person to a question. The specific person may be, for example, a real person. The specific person may also be, for example, a fictional person.

[0017] In Figure 1, the information processing device 100 can utilize a first language model 110, a second language model 120, and a third language model 130. The first language model 110 has the function of outputting text as a response to input text. The second language model 120 has the function of outputting text as a response to input text. The third language model 130 has the function of outputting text as a response to input text.

[0018] The first language model 110, the second language model 120, and the third language model 130 are, for example, the same LLM. The LLM may be, for example, a Character-LLM. The first language model 110, the second language model 120, and the third language model 130 may each be, for example, different LLMs. The information processing device 100 has, for example, the first language model 110, the second language model 120, and the third language model 130.

[0019] Furthermore, the information processing device 100 acquires persona information 140. The persona information 140 includes information 141 for each of one or more items relating to the characteristics of the subject. The characteristics of the subject are, for example, the personality of the subject. For example, the persona information 140 includes information 141 for each of one or more items defining the personality of the subject.

[0020] One or more items include, for example, items relating to attributes that define the personality of the subject. Attributes include, for example, name, gender, age, occupation, or address. One or more items include, for example, items relating to a profile or life story of a specific person who possesses the personality of the subject. The profile includes, for example, beliefs. Beliefs specifically refer to views on privacy. The information processing device 100 acquires persona information 140 by accepting input of persona information 140 based on user operation input, for example.

[0021] (1-1) The information processing device 100 acquires question information 150. Question information 150 is information relating to a question. Question information 150 represents a question. Question information 150 includes, for example, text representing the content of the question. The information processing device 100 acquires question information 150 by, for example, accepting input of question information 150 based on user operation input.

[0022] (1-2) The information processing device 100 uses the first language model 110 to generate first information 111 that explains the content of the acquired question information 150 that relates to the acquired persona information 140. The content related to the persona information 140 is, for example, the part of the question information 150 that was evaluated as having high importance according to the persona information 140.

[0023] The information processing device 100 generates a first prompt, for example, which includes question information 150 and requests a description of content related to persona information 140. The information processing device 100 generates first information 111 based on the text that the first language model 110 outputs in response to the input of the generated first prompt to the first language model 110. The information processing device 100 may also acquire the first information 111 generated using the first language model 110. The information processing device 100 may also generate the first information 111 by other methods that do not utilize the first language model 110.

[0024] (1-3) The information processing device 100 uses the second language model 120 to generate second information 121 that explains the meaning of the acquired question information 150. The information processing device 100 generates, for example, a second prompt that includes the question information 150 and requests an explanation of the meaning of the question information 150. The information processing device 100 generates second information 121 based on the text that is the response output by the second language model 120 in response to the input of the generated second prompt to the second language model 120. The information processing device 100 may acquire the second information 121 generated using the second language model 120. The information processing device 100 may generate second information 121 by other methods that do not utilize the second language model 120.

[0025] (1-4) The information processing device 100 uses the third language model 130 to generate third information 131 that explains the response to a question, corresponding to the information 141 of one or more items, based on the generated first information 111 and the generated second information 121. The information processing device 100 generates a third prompt that includes the first information 111 and the second information 121 and requests an explanation of the response to a question, corresponding to the information 141 of each item. The information processing device 100 generates third information 131 based on the text that is the response output by the third language model 130 in response to the generated third prompt being input to the third language model 130. The information processing device 100 may also obtain the third information 131 generated using the third language model 130. The information processing device 100 may generate third information 131 by other methods that do not utilize the third language model 130.

[0026] (1-5) The information processing device 100 generates response information 160 to the question information 150 based on the generated third information 131. The response information 160 represents text that is a response to the question information 150 by a specific person possessing the target personality. For example, the information processing device 100 generates response information 160 by summarizing the generated third information 131. As a result, the information processing device 100 can accurately generate response information 160 that appropriately reflects the persona information 140 and mimics a response to the question information 150 by a specific person possessing the target personality.

[0027] The information processing device 100 may utilize one or more language models that output text as a response to input text for at least one of the processes among the process of generating first information 111, the process of generating second information 121, the process of generating third information 131, and the process of generating response information 160.

[0028] The information processing device 100 can generate response information 160 so as to comprehensively include third information that explains the response to a question, corresponding to the information 141 of each item that makes up the persona information 140. Specifically, when generating the response information 160, the information processing device 100 can comprehensively consider the information 141 of each item that makes up the persona information 140. Specifically, after comprehensively considering the information 141 of each item, the information processing device 100 can generate response information 160 that accurately mimics the response of a specific person with the target personality to the question information 150.

[0029] Here, we have described a case in which the information processing device 100 has a first language model 110, a second language model 120, and a third language model 130, but it is not limited to this. For example, the information processing device 100 may be able to use the first language model 110 by communicating with another computer that has the first language model 110.

[0030] Furthermore, for example, the information processing device 100 may be able to utilize the second language model 120 by communicating with another computer having the second language model 120. Also, for example, the information processing device 100 may be able to utilize the third language model 130 by communicating with another computer having the third language model 130.

[0031] Here, we have described a case in which the information processing device 100 acquires persona information 140 by receiving the persona information 140 based on user input, but it is not limited to this. For example, the information processing device 100 may acquire persona information 140 by receiving it from another computer.

[0032] This explanation describes a case in which the information processing device 100 acquires question information 150 by receiving the question information 150 based on user input, but it is not limited to this. For example, the information processing device 100 may acquire question information 150 by receiving it from another computer.

[0033] Here, we have described the case where the functions of the information processing device 100 are realized by a single computer, but this is not the only case. For example, the functions of the information processing device 100 may be realized through the collaboration of multiple computers. For example, the functions of the information processing device 100 may be realized on the cloud.

[0034] (An example of an information processing system 200) Next, an example of an information processing system 200 that applies the information processing device 100 shown in Figure 1 will be described using Figure 2.

[0035] Figure 2 is an explanatory diagram showing an example of an information processing system 200. In Figure 2, the information processing system 200 includes a control device 201, a language processing device 202, and one or more client devices 203.

[0036] In the information processing system 200, the control device 201 and the language processing device 202 are connected via a wired or wireless network 210. The network 210 is, for example, a LAN (Local Area Network), a WAN (Wide Area Network), or the Internet. Also in the information processing system 200, the control device 201 and the client device 203 are connected via a wired or wireless network 210.

[0037] The control device 201 is a computer designed to mimic the responses of a specific person to a question. The control device 201 operates as an information processing device 100. The control device 201 acquires persona information. The control device 201 acquires persona information, for example, by accepting input of persona information based on user input.

[0038] Persona information includes information for each of several items relating to the characteristics of the subject. For example, persona information includes information for each of several items defining the personality of the subject. These items may include, for example, items relating to attributes that define the personality of the subject. These items may include, for example, items relating to the profile or life story of a specific person who possesses the personality of the subject. The profile may include, for example, beliefs. These beliefs may specifically include views on privacy.

[0039] The control device 201 acquires question information relating to questions directed at a specific person possessing the target personality. The control device 201 may acquire question information, for example, by receiving the question information from the client device 203. Based on the acquired question information and the acquired persona information, the control device 201 may extract information for one or more items that are evaluated as having a high degree of relevance to the question from among the information for each item of a plurality of items. The control device 201 may newly acquire persona information that includes the information for each of the one or more items that have been extracted.

[0040] The control device 201, upon receiving question information, generates response information that mimics the response of a specific person possessing the target personality to the question. For example, in cooperation with the language processing device 202, the control device 201 generates response information that mimics the response of a specific person possessing the target personality to the question, based on the acquired persona information, similar to the case in Figure 1.

[0041] The control device 201 outputs the generated response information. The output format may be, for example, display on a screen, print to a printer, transmit to another computer, or store in a memory area. The control device 201 transmits the generated response information to the client device 203, for example, the control device 201 is a server or a PC.

[0042] The language processing device 202 is a computer that executes an LLM. The LLM may be, for example, a Character-LLM. The language processing device 202 executes the LLM according to the control of the control device 201. The language processing device 202, for example, receives a prompt for the LLM from the control device 201. The language processing device 202, for example, transmits text that is a response output by the LLM in response to inputting the received prompt to the LLM to the control device 201. The language processing device 202 is, for example, a server or a PC or the like.

[0043] The client device 203 is a computer that generates question information regarding a question. The client device 203 is used by a user who requests a response from a specific person to the question. The client device 203 generates question information regarding a question based on a user's operation input. The client device 203 transmits the generated question information to the information processing device 100.

[0044] The client device 203 receives response information that mimics a response by a specific person to a question from the information processing device 100. The client device 203 outputs the received response information so that the user can refer to it. The client device 203 is, for example, a PC, a tablet terminal, or a smartphone or the like.

[0045] Here, the case where the control device 201 is a device different from the language processing device 202 has been described, but it is not limited to this. For example, the control device 201 may have a function as the language processing device 202 and may also operate as the language processing device 202. In this case, the information processing system 200 may not include the language processing device 202.

[0046] Here, the case where the control device 201 is a device different from the client device 203 has been described, but it is not limited to this. For example, the control device 201 may have a function as the client device 203 and may also operate as the client device 203. In this case, the information processing system 200 may not include the client device 203.

[0047] (Hardware Configuration Example of Control Device 201) Next, a hardware configuration example of the control device 201 will be described using FIG. 3.

[0048] FIG. 3 is a block diagram showing a hardware configuration example of the control device 201. In FIG. 3, the control device 201 includes a CPU (Central Processing Unit) 301, a memory 302, and a network I / F (Interface) 303. Further, the control device 201 includes a recording medium I / F 304, a recording medium 305, a display 306, and an input device 307. Also, each component is connected by a bus 300.

[0049] Here, the CPU 301 controls the overall operation of the control device 201. The memory 302 includes, for example, a ROM (Read Only Memory), a RAM (Random Access Memory), and a flash ROM. Specifically, for example, the flash ROM and the ROM store various programs, and the RAM is used as the work area of the CPU 301. The programs stored in the memory 302 are loaded into the CPU 301 to cause the CPU 301 to execute the coded processes.

[0050] The network I / F 303 is connected to the network 210 through a communication line and is connected to other computers via the network 210. Then, the network I / F 303 manages the interface between the network 210 and the internal components and controls the input / output of data from other computers. The network I / F 303 is, for example, a modem or a LAN adapter.

[0051] The recording medium interface 304 controls the reading and writing of data to the recording medium 305 according to the control of the CPU 301. The recording medium interface 304 is, for example, a disk drive, an SSD (Solid State Drive), or a USB (Universal Serial Bus) port. The recording medium 305 is a non-volatile memory that stores the data written under the control of the recording medium interface 304. The recording medium 305 is, for example, a disk, semiconductor memory, or USB memory. The recording medium 305 may be detachable from the control device 201.

[0052] Display 306 displays data such as cursors, icons, toolboxes, documents, images, or functional information. Display 306 is, for example, a CRT (Cathode Ray Tube), a liquid crystal display, or an organic EL (Electroluminescence) display. Input device 307 has keys for inputting characters, numbers, or various instructions, and performs data input. Input device 307 is, for example, a keyboard or a mouse. Input device 307 may also be, for example, a touch panel input pad or a numeric keypad.

[0053] The control device 201 may have, in addition to the components described above, a camera, for example. Furthermore, the control device 201 may have, in addition to the components described above, a printer, scanner, microphone, or speaker, for example. Also, the control device 201 may have multiple recording medium interfaces 304 and recording media 305, for example. Furthermore, the control device 201 does not necessarily have, for example, a display 306 or an input device 307. Also, the control device 201 does not necessarily have, for example, recording medium interfaces 304 and recording media 305.

[0054] (Example of hardware configuration of language processing device 202) The example of hardware configuration of language processing device 202 is specifically the same as the example of hardware configuration of control device 201 shown in Figure 3, so the explanation is omitted.

[0055] (Example of hardware configuration of client device 203) The example of hardware configuration of client device 203 is specifically the same as the example of hardware configuration of control device 201 shown in Figure 3, so the explanation is omitted.

[0056] (Example of Functional Configuration of Information Processing Device 100) Next, an example of the functional configuration of the information processing device 100 will be described using Figure 4.

[0057] Figure 4 is a block diagram showing an example of the functional configuration of the information processing device 100. The information processing device 100 includes a storage unit 400, an acquisition unit 401, an extraction unit 402, a first generation unit 403, a second generation unit 404, a third generation unit 405, a response generation unit 406, and an output unit 407.

[0058] The storage unit 400 is implemented by a storage area such as the memory 302 or recording medium 305 shown in Figure 3. The following description will focus on the case where the storage unit 400 is included in the information processing device 100, but is not limited to this case. For example, the storage unit 400 may be included in a device different from the information processing device 100, and the contents of the storage unit 400 may be accessible from the information processing device 100.

[0059] The acquisition unit 401 to the output unit 407 function as an example of a control unit. Specifically, the acquisition unit 401 to the output unit 407 realize their functions, for example, by having the CPU 301 execute a program stored in a storage area such as the memory 302 or recording medium 305 shown in Figure 3, or by using the network interface 303. The processing results of each functional unit are stored in a storage area such as the memory 302 or recording medium 305 shown in Figure 3.

[0060] The storage unit 400 stores various types of information that are referenced or updated during the processing of each functional unit. For example, the storage unit 400 stores question information. The question information represents the content of a question. The question information includes, for example, text that represents the content of the question. The question information is acquired, for example, by the acquisition unit 401.

[0061] The memory unit 400 stores, for example, persona information. Persona information includes, for example, information for each of one or more items relating to the characteristics of the subject. Persona information includes, for example, information for each of one or more items defining the personality of the subject. Items represent, for example, attributes that define the personality of the subject. Attributes are, for example, name, gender, age, occupation, or address. Items may also represent, for example, a profile or life story of a specific person who possesses the personality of the subject. Profiles include, for example, beliefs. Beliefs specifically include beliefs such as views on privacy. Persona information specifically includes profile or life story information of a specific person corresponding to the personality of the subject. Persona information is acquired, for example, by the acquisition unit 401. Persona information may be pre-set by the user, for example.

[0062] The memory unit 400 may store, for example, language models. The language model has the function of outputting text that is a response to input text. The language model is, for example, LLM. Specifically, the memory unit 400 may store a first language model, a second language model, a third language model, a fourth language model, and a fifth language model.

[0063] Here, the first language model, the second language model, the third language model, the fourth language model, and the fifth language model may be the same LLM. Alternatively, the first language model, the second language model, the third language model, the fourth language model, and the fifth language model may each be separate LLMs. The storage unit 400 may store one or more language models.

[0064] The first language model is used, for example, to explain the parts of the question information that are related to persona information. The parts related to persona information are, for example, the parts of the question information that have been evaluated as highly important according to the persona information. The first language model may also be an LLM that has been fine-tuned to improve the accuracy of explaining the parts of the question information that have been evaluated as highly important.

[0065] The second language model is used, for example, to explain the meaning of question information. The second language model may be a finely tuned LLM to improve the accuracy of explaining the meaning of question information.

[0066] A third language model (LLM) is used, for example, to explain responses to questions based on information from one or more items related to the characteristics of a subject. The LLM may be a finely tuned LLM to improve the accuracy of explaining responses to questions based on information from each item.

[0067] The fourth language model is used, for example, to extract information from one or more items that are evaluated as having a high degree of relevance to a question, from among the information of each item of multiple items related to the characteristics of the subject. The fourth language model may also be an LLM that has been fine-tuned to improve the accuracy of extracting information from one or more items that are evaluated as having a high degree of relevance to a question.

[0068] The fifth language model is used, for example, to mimic the response of a specific person possessing the target personality to a question. The fifth language model may be a finely tuned LLM to improve the accuracy of mimicking the response of a specific person possessing the target personality to a question. The language model is pre-configured, for example, by the user. The language model may be acquired, for example, by the acquisition unit 401.

[0069] Here, the memory unit 400 may not store the language model. For example, the information processing device 100 may utilize the language model by communicating with another computer that has a language model. The other computer is, for example, the language processing device 202.

[0070] The acquisition unit 401 acquires various types of information used in the processing of each functional unit. The acquisition unit 401 stores the acquired information in the storage unit 400 or outputs it to each functional unit. The acquisition unit 401 may also output the information stored in the storage unit 400 to each functional unit. The acquisition unit 401 acquires various types of information, for example, based on user input. The acquisition unit 401 may also receive various types of information from a device other than the information processing device 100, for example.

[0071] The acquisition unit 401 acquires, for example, question information. Specifically, the acquisition unit 401 acquires question information by receiving input of question information. Specifically, the acquisition unit 401 may acquire question information by receiving question information from another computer. The other computer is, for example, a client device 203.

[0072] The acquisition unit 401 acquires, for example, persona information. Specifically, the acquisition unit 401 acquires persona information by receiving persona information as input. Specifically, the acquisition unit 401 may acquire persona information by receiving persona information from another computer. The other computer is, for example, a client device 203.

[0073] The acquisition unit 401 acquires, for example, a first language model. Specifically, the acquisition unit 401 acquires the first language model by receiving input of the first language model. Specifically, the acquisition unit 401 may acquire the first language model by receiving the first language model from another computer. The other computer is, for example, a client device 203.

[0074] The acquisition unit 401 acquires, for example, a second language model. Specifically, the acquisition unit 401 acquires a second language model by receiving input of a second language model. Specifically, the acquisition unit 401 may acquire a second language model by receiving a second language model from another computer. The other computer is, for example, a client device 203.

[0075] The acquisition unit 401 acquires, for example, a third language model. Specifically, the acquisition unit 401 acquires a third language model by receiving input of a third language model. Specifically, the acquisition unit 401 may acquire a third language model by receiving a third language model from another computer. The other computer is, for example, a client device 203.

[0076] The acquisition unit 401 acquires, for example, a fourth language model. Specifically, the acquisition unit 401 acquires a fourth language model by receiving input of a fourth language model. Specifically, the acquisition unit 401 may acquire a fourth language model by receiving a fourth language model from another computer. The other computer is, for example, a client device 203.

[0077] The acquisition unit 401 acquires, for example, a fifth language model. Specifically, the acquisition unit 401 acquires a fifth language model by receiving input of a fifth language model. Specifically, the acquisition unit 401 may acquire a fifth language model by receiving a fifth language model from another computer. The other computer is, for example, a client device 203.

[0078] The acquisition unit 401 may receive a start trigger to initiate processing in any of the functional units. A start trigger may be, for example, a predetermined operation input by the user. A start trigger may also be, for example, the reception of predetermined information from another computer. A start trigger may also be, for example, the output of predetermined information by any of the functional units. The acquisition unit 401 accepts, for example, the acquisition of question information as a start trigger to initiate processing in the extraction unit 402, the first generation unit 403, the second generation unit 404, the third generation unit 405, and the response generation unit 406.

[0079] The extraction unit 402 uses a fourth language model to extract information from one or more items that are evaluated as having a high degree of relevance to the question, based on the acquired question information. The extraction unit 402 then acquires new persona information that includes the information from each of the one or more extracted items.

[0080] The extraction unit 402 generates a prompt that includes question information and requests the extraction of information for one or more items that are evaluated as having a high degree of relevance to the question from among the information for each of the multiple items relating to the characteristics of the target. The extraction unit 402 obtains text that will be the response output by the fourth language model in response to the input of the generated prompt to the fourth language model. Based on the obtained text, the extraction unit 402 identifies the information for one or more items that are evaluated as having a high degree of relevance to the question, extracted from the information for each of the multiple items, and obtains new persona information.

[0081] This allows the extraction unit 402 to identify which items of information it is preferable to consider when mimicking the response of a specific person possessing the target personality to a question. Therefore, even when the persona information acquired by the acquisition unit 401 contains a large amount of information on various items, the extraction unit 402 can improve the efficiency of the process of mimicking the response of a specific person possessing the target personality to a question.

[0082] The first generation unit 403 uses the first language model to generate first information that explains the content of the acquired question information that is related to the persona information acquired by the acquisition unit 401 or the extraction unit 402. The content related to the persona information is, for example, the parts of the question information that were evaluated as having high importance according to the persona information.

[0083] The first generation unit 403 generates a prompt that includes, for example, question information and persona information, and requests an explanation of the content of the question information related to the persona information. The first generation unit 403 obtains, for example, text that will be the response output by the first language model in response to the input of the generated prompt to the first language model. The first generation unit 403 generates, for example, first information based on the obtained text. The first information may be, for example, the text output by the first language model itself. In this way, the first generation unit 403 can obtain first information that serves as a guideline for imitating the response of a specific person with the target personality to a question.

[0084] The second generation unit 404 uses the second language model to generate second information that explains the meaning of the acquired question information. The second generation unit 404 generates a prompt, for example, that includes the question information and requests an explanation of the meaning of the question information. The second generation unit 404 obtains text that will be the response output by the second language model when the generated prompt is input to the second language model. The second generation unit 404 generates second information based on the obtained text. The second information may be, for example, the text output by the second language model itself. In this way, the second generation unit 404 can obtain second information that serves as a guideline for mimicking the response of a specific person with the target personality to a question.

[0085] The third generation unit 405 uses a third language model to generate third information corresponding to the information of each of one or more items, based on the persona information acquired by the acquisition unit 401 or the extraction unit 402, the generated first information, and the generated second information. The third information explains the response to the question. The third information can, for example, be an element that forms the response to the question. The third information is generated, for example, for each item of information.

[0086] The third generation unit 405 generates a prompt that requests the generation of third information corresponding to the information of at least one of the items, which includes, for example, persona information, first information, and second information. The third generation unit 405 obtains, for example, the text that would be the response output by the third language model when the generated prompt is input to the third language model. Based on the obtained text, the third generation unit 405 generates third information corresponding to the information of at least one of the items. The third information may be, for example, the text output by the third language model itself.

[0087] The third generation unit 405 repeatedly generates third information corresponding to the information of at least one of the items, as described above, until it has collected third information corresponding to the information of each of the items. This allows the third generation unit 405 to comprehensively consider the information of each of the items and collect third information that forms the elements of a response from a specific person with the target personality to a question, making it easier to generate response information with high accuracy. Furthermore, since the third generation unit 405 can associate third information with the information of each item and generate third information corresponding to that item, it can provide the basis for the response.

[0088] The response generation unit 406 generates response information to the question information based on the generated third information. The response information mimics the response of a specific person possessing the target personality to the question. The response generation unit 406, for example, uses a fifth language model to generate response information that summarizes and explains the generated third information according to the characteristics of the target. The characteristics of the target are, for example, the target personality. The response information mimics, for example, the tone of voice of a specific person possessing the target personality and summarizes and explains the generated third information.

[0089] Specifically, the response generation unit 406 generates a prompt that includes the generated third information and requests a summary explanation of the generated third information according to the characteristics of the target. The characteristics of the target are, for example, the target's personality. Specifically, the response generation unit 406 obtains text that will be the response output by the fifth language model in response to the input of the generated prompt to the fifth language model. Specifically, the response generation unit 406 generates response information based on the obtained text. The response information may be the text output by the fifth language model itself. In this way, the response generation unit 406 can accurately mimic the response of a specific person with the target's personality to a question.

[0090] Furthermore, the response information may specifically represent the content of the elements that form the response, based on third information corresponding to the information of each item. This allows the response generation unit 406 to generate response information that can present the basis for each element that forms the response.

[0091] The output unit 407 outputs the processing result of at least one of the functional units. The output format can be, for example, display on a screen, print to a printer, transmit to an external device via the network interface 303, or store in a storage area such as the memory 302 or recording medium 305. This allows the output unit 407 to notify the user of the processing result of at least one of the functional units, thereby improving the usability of the information processing device 100.

[0092] The output unit 407 outputs, for example, the response information generated by the response generation unit 406. Specifically, the output unit 407 outputs the response information so that the user can refer to it. Specifically, the output unit 407 may also transmit the response information to another computer. The other computer is, for example, the client device 203.

[0093] Here, we have described a case in which the information processing device 100 sequentially generates prompts requesting the generation of first information, second information, third information, and response information, but it is not limited to this case. For example, the information processing device 100 may generate a single prompt that requests the sequential generation of first information, second information, third information, and response information. In this case, the language model generates first information, second information, third information, and response information sequentially according to the prompt.

[0094] (Example of operation of the information processing device 100) Next, an example of operation of the information processing device 100 will be explained using Figures 5 to 11. First, using Figure 5, an example of persona information 500, which is stored by the information processing device 100 and includes information on each of the multiple items that define the target persona, will be explained.

[0095] Figure 5 is an explanatory diagram showing an example of persona information 500. In Figure 5, persona information 500 includes information for each of several items that define the target persona. The items include, for example, name, gender, age, occupation, address, personality, hobbies, and views and feelings about privacy. Other items include, for example, internet banking usage, specific usage and frequency of main and sub-banks, frequency of use and views on convenience store ATMs, and specific opinions and impressions on continuous customer management.

[0096] The information processing device 100 acquires and stores persona information 500. The information processing device 100 acquires persona information 500 by receiving input of persona information 500 based on user operation input, for example. The information processing device 100 may also acquire persona information 500 by receiving persona information 500 from the client device 203.

[0097] Next, using Figures 6 to 11, an example will be described in which the information processing device 100 generates response information 1120 that mimics the response of a specific person possessing the target persona to a question, based on the persona information 500. In the example shown in Figures 6 to 11, the information processing device 100 can utilize the LLM 600.

[0098] Specifically, the information processing device 100 generates response information 1120 by using the LLM 600 to perform the process of step 1, which will be described later, as shown in Figures 6 and 7, and the process of step 2, which will be described later, as shown in Figures 8 to 11. First, an example of the process of step 1 will be explained using Figures 6 and 7.

[0099] Figures 6 and 7 are explanatory diagrams showing an example of the processing in step 1. In Figure 6, the information processing device 100 acquires question information 610 representing the content of the question. The information processing device 100 acquires the question information 610, for example, by receiving the question information 610 from the client device 203. Here, it is considered preferable to reflect whether or not the persona information 500 is used for internet banking in the response to the question.

[0100] Based on the question information 610, the information processing device 100 extracts information for each of one or more items from the persona information 500 that are highly relevant to the question and are considered important when generating a response to the question. For example, the information processing device 100 generates a prompt that includes the question information 610 and the persona information 500, and requests the extraction of information for each of one or more items from the persona information 500.

[0101] Specifically, the information processing device 100 generates a prompt by inserting the question information 610 and the persona information 500 into the prompt template 700. Now, we will move on to the explanation of Figure 7 and describe an example of the prompt template 700.

[0102] As shown in Figure 7, the template 700 includes, for example, a section 701 that requests the LLM 600 to engage in dialogue tailored to the target persona, based on the persona information 500. The template 700 also includes, for example, a section 702 for inserting the persona information 500.

[0103] Template 700 includes, for example, a section 703 that requests LLM 600 to refer to question information 610 and extract information for each of the one or more items from persona information 500 that are evaluated as having a high degree of relevance to the question. Template 700 includes, for example, a section 704 for inserting question information 610. Template 700 includes, for example, a section 705 for describing definitions of terms. Template 700 includes, for example, a section 706 for describing example responses to prompts.

[0104] Returning to the explanation of Figure 6, the information processing device 100 acquires information for each of the one or more extracted items based on the text response output by the LLM 600 in response to the input of the generated prompt to the LLM 600. The information processing device 100 generates extracted persona information 620 that includes the information for each of the one or more acquired items.

[0105] As a result, even when there is a large amount of information on the items included in the persona information 500, the information processing device 100 can narrow down the information on the items to be referenced in order to improve the efficiency of the process of imiting the response of a specific person with the target personality to a question. In addition, by narrowing down the information on the items to be referenced, the information processing device 100 can make it less likely for the LLM 600 to miss information that is relatively important. Next, an example of the processing in step 2 will be explained using Figures 8 to 11.

[0106] Figures 8 to 11 are explanatory diagrams showing an example of the processing in step 2. In Figure 8, the information processing device 100 sequentially generates response information 1120 that mimics the response of a specific person possessing the target persona to a question, based on the question information 610 and the extracted persona information 620. Specifically, the information processing device 100 generates a prompt by inserting the question information 610 and the extracted persona information 620 into the prompt template 800.

[0107] As shown in Figure 8, the template 800 includes, for example, a section 801 that requests the LLM 600 to engage in dialogue tailored to the target persona, based on the extracted persona information 620. The template 800 also includes, for example, a section 802 for inserting the extracted persona information 620. The template 800 also includes, for example, a section 803 for inserting reference information. The reference information defines, for example, terms that appear in the question information 610.

[0108] Template 800 includes, for example, a section 804 that describes the rules for responses. Template 800 includes, for example, a section 805 that requests LLM 600 to generate response information 1120 for question information 610. Section 805 includes, for example, a section for inserting question information 610. Template 800 includes, for example, a section 806 that describes the procedure for LLM 600 to generate response information 1120. Template 800 includes, for example, a section 807 that describes an example response to a prompt.

[0109] The procedure includes, for example, the first phase, the Attention phase. The procedure includes, for example, the second phase, the Comprehension phase. The procedure includes, for example, the third phase, the Thinking phase. The procedure includes, for example, the fourth phase, the Consolidation of Thoughts phase. The procedure includes, for example, the fifth phase, the Check Thoughts phase. The procedure includes, for example, the sixth phase, the Summary of Thoughts phase. The procedure includes, for example, the seventh phase, the Language Production phase.

[0110] The Attention phase is, for example, the phase in which you explain which parts of the question you evaluated as being important and why those parts are important. The Comprehension phase is, for example, the phase in which you interpret the intent of the question. The Thinking phase is the phase in which you interpret the information of each item of one or more items. In the Thinking phase, for example, you interpret the information of each item by tagging it.

[0111] The Consolidation of Thoughts phase is a phase in which, based on the interpretation of the information for each item, the elements that form the answer to the question are explained for each item's information. In the Consolidation of Thoughts phase, for each item's information, the elements that form the answer to the question are explained while clearly indicating the tags assigned to that item's information. The Check Thoughts phase is a phase in which it is confirmed whether the elements that form the answer to the question have been comprehensively explained for all items' information.

[0112] The Summary of Thoughts phase is the phase in which the elements that form the answers to the explained questions regarding the information of each item are compiled. The Language Production phase is the phase in which the compiled elements are verbalized in a way that suits the persona, thereby generating the answers to the questions.

[0113] The information processing device 100 generates response information 1120 based on the text output by the LLM 600 in response to the input of the generated prompt to the LLM 600. The response information 1120 may be the text output by the LLM 600 itself. In this way, the information processing device 100 can cause the LLM 600 to execute each phase that forms the procedure for generating the response information 1120, which is the processing in step 2. Now, we will move on to the explanation of Figure 9 and describe an example of the operation of the LLM 600.

[0114] In Figure 9, during the Attention phase, the LLM 600 generates first information 910 that explains the parts of the question that were evaluated as highly important and the reasons why those parts were highly important, based on the question information 610 and the extracted persona information 620. During the Comprehension phase, the LLM 600 generates second information 920 that interprets the intent of the question based on the question information 610. Next, we will move on to the explanation of Figure 10.

[0115] In Figure 10, during the Thinking phase, the LLM 600 generates third-party information 1010 by interpreting the information for each of one or more items based on the question information 610 and the extracted persona information 620. For example, the LLM 600 generates third-party information 1010 with the tag "No. x" attached to the information for each of one or more items.

[0116] In the Consolidation of Thoughts phase, LLM600 generates fourth information 1020 that includes elements forming a response to a question with respect to information on at least one of the items. For example, LLM600 generates fourth information 1020 that includes elements relating to information on at least one of the items based on question information 610, extracted persona information 620, first information 910, second information 920, and third information 1010. Specifically, LLM600 generates fourth information 1020 that describes elements relating to information on any of the items, explicitly indicating the tags assigned to the information on that item.

[0117] In the Check Thoughts phase, LLM600 determines whether elements related to the information of all items exist in the fourth information 1020. If elements related to the information of any item do not exist in the fourth information 1020, LLM600 returns to the Consolidation of Thoughts phase and regenerates the fourth information 1020 to include elements related to the information of that item.

[0118] If elements related to the information of all items are present in the fourth information 1020, the LLM600 outputs text 1030 indicating that elements related to the information of all items are present, and then proceeds to the Summary of Thoughts phase. In the Summary of Thoughts phase, the LLM600 generates fifth information 1040 which summarizes the elements related to the information of each item. Next, we will proceed to the explanation of Figure 11.

[0119] In Figure 11, the LLM 600 generates response information 1120 in the Language Production phase by verbalizing the fifth piece of information 1040 to match the persona. As a result, the information processing device 100 can use the LLM 600 to generate response information 1120 that comprehensively considers the information for each item and accurately mimics the response of a specific person corresponding to the persona to the question.

[0120] The information processing device 100 can generate response information 1120 that appropriately takes into account, for example, whether or not internet banking is used. Furthermore, the information processing device 100 can generate fourth information 1020 that explains the elements of the information for each item, explicitly indicating the tags assigned to the information for that item. Therefore, the information processing device 100 can present the basis for each element that forms the response.

[0121] On the other hand, conventionally, LLM600 tends to generate response information 1110 based on, for example, persona information 500 and question information 610, without comprehensively considering the information for each item. Thus, conventionally, response information 1110 may be generated without considering information regarding internet banking usage. Also, people with a high awareness of privacy tend to avoid using convenience store ATMs. However, conventionally, response information 1110 may be generated that contradicts the user's awareness of privacy and their tendency to use convenience store ATMs.

[0122] In contrast, the proposed method by the information processing device 100 can generate consistent response information 1120 for the LLM 600 that comprehensively considers the information for each item. Since the proposed method by the information processing device 100 controls the LLM 600 with prompts, it is not necessary to train the LLM 600 with persona information 500. Therefore, the proposed method by the information processing device 100 can reduce the processing burden when generating response information 1110 for multiple pieces of persona information 500.

[0123] Here, we have described a case in which the information processing device 100 inputs the generated prompts to the LLM 600 to have the LLM 600 execute each phase that forms the procedure for generating the response information 1120 which is the process for step 2, all at once, but this is not limited to this case. For example, the information processing device 100 may generate a prompt for each phase requesting that the relevant phase be executed, and input the generated prompts to the LLM 600 to have the LLM 600 execute each phase sequentially.

[0124] (Overall Processing Procedure) Next, an example of the overall processing procedure executed by the information processing device 100 will be described using Figure 12. The overall processing is realized, for example, by the CPU 301 shown in Figure 3, storage areas such as memory 302 and recording medium 305, and network I / F 303.

[0125] Figure 12 is a flowchart showing an example of the overall processing procedure. In Figure 12, the information processing device 100 acquires question information (step S1201). Next, the information processing device 100 acquires persona information about the target personality (step S1202). Then, the information processing device 100 uses LLM to extract information for each of the one or more items from the persona information that are evaluated as having a high degree of relevance to the question (step S1203).

[0126] Next, the information processing device 100 uses LLM to generate first information that explains the parts of the acquired question information that were evaluated as having high importance, based on the information of each of the one or more extracted items (step S1204). Then, the information processing device 100 uses LLM to generate second information that explains the meaning of the acquired question information (step S1205).

[0127] Next, the information processing device 100 uses LLM to perform a generation process, which will be described later in Figure 13, and generates response information that mimics a response to a question based on the generated first information and the generated second information (step S1206). Then, the information processing device 100 uses LLM to convert the response information to match the tone of voice corresponding to the target person (step S1207).

[0128] Next, the information processing device 100 outputs the converted response information (step S1208). Then, the information processing device 100 terminates the entire process. As a result, the information processing device 100 can generate response information that accurately mimics the response of a specific person possessing the target personality in response to a question.

[0129] (Generation Processing Procedure) Next, an example of a generation processing procedure executed by the information processing device 100 will be described using Figure 13. The generation processing is realized, for example, by the CPU 301 shown in Figure 3, a storage area such as the memory 302 and recording medium 305, and the network I / F 303.

[0130] Figure 13 is a flowchart showing an example of the generation process procedure. In Figure 13, the information processing device 100 uses LLM to tag the information of each of the one or more extracted items based on the first information and the second information, and generates tag information that describes the information of each tagged item (step S1301).

[0131] Next, the information processing device 100 uses LLM to generate third information that, based on the generated tag information, explains the elements that form a response to a question, corresponding to the information of the item to which each tag is assigned, while explicitly indicating the tag (step S1302).

[0132] Next, the information processing device 100 uses LLM to determine whether all the elements necessary to form a response to the question, corresponding to the information of all items, are available (step S1303). If there are insufficient elements to form a response to the question (step S1303: No), the information processing device 100 returns to the process in step S1302. On the other hand, if all the elements necessary to form a response to the question are available (step S1303: Yes), the information processing device 100 uses LLM to generate response information that combines the generated third information (step S1304). After that, the information processing device 100 terminates the generation process.

[0133] Here, the information processing device 100 may execute some steps of the flowcharts in Figures 12 and 13 in a different order. For example, the order of steps S1204 and S1205 can be changed. Also, the information processing device 100 may omit some steps of the flowcharts in Figures 12 and 13. For example, the process in step S1303 can be omitted.

[0134] (Examples of application of the information processing device 100) For example, the information processing device 100 is applied when realizing a service that presents response information that mimics the response of a specific person to a question. Specifically, the information processing device 100 is applied when realizing a service that supports employee training related to sales by presenting response information that mimics the response of a person at a sales destination to a question.

[0135] More specifically, the information processing device 100 may be applied to a service that supports market research on a specific product by presenting response information that mimics customer responses to questions about that specific product. More specifically, the information processing device 100 may be applied to a service that supports a service center by presenting response information that mimics responses from service center staff to questions.

[0136] As explained above, the information processing device 100 can acquire question information related to a question. The information processing device 100 can acquire persona information that includes information on each of one or more items related to the characteristics of the target. The information processing device 100 can generate first information that explains the content related to the acquired persona information from the acquired question information. The information processing device 100 can generate second information that explains the meaning of the acquired question information. Based on the generated first information and the generated second information, the information processing device 100 can generate third information that explains the response to the question according to the information on each of the one or more items. Based on the generated third information, the information processing device 100 can generate response information to the question information. As a result, the information processing device 100 can generate response information with high accuracy.

[0137] According to the information processing device 100, one or more language models can be used for at least one of the processes among the process of generating first information, the process of generating second information, the process of generating third information, and the process of generating response information. This makes it easier for the information processing device 100 to implement at least one process using a language model.

[0138] According to the information processing device 100, using a language model, it is possible to extract information on one or more items that are evaluated as having a high degree of relevance to the question from among the information on multiple items related to the characteristics of the target, based on the acquired question information. According to the information processing device 100, it is possible to acquire persona information that includes the information on each of the one or more extracted items. As a result, the information processing device 100 can extract information on one or more items that are appropriate from the standpoint of generating response information, thereby improving the accuracy of generating response information.

[0139] According to the information processing device 100, a prompt can be generated that includes the generated first information and the generated second information, and requests third information corresponding to the information of at least one of one or more items. According to the information processing device 100, when the generated prompt is input to a language model, the text output by the language model can be obtained. According to the information processing device 100, based on the obtained text, third information corresponding to the information of at least one of one or more items can be generated. According to the information processing device 100, the generation of third information can be repeated until third information corresponding to the information of each of the one or more items is obtained. In this way, the information processing device 100 can obtain third information corresponding to the information of each of the one or more items.

[0140] According to the information processing device 100, it is possible to generate response information that summarizes and explains the generated third information according to the characteristics of the target, using a language model. This makes it easier for the user to understand the response information.

[0141] According to the information processing device 100, persona information, including profile or life story information, can be acquired for a specific person corresponding to a personality. This allows the information processing device 100 to improve the accuracy of generating response information by utilizing the persona information, which contains useful information.

[0142] The information processing method described in this embodiment can be implemented by executing a pre-prepared program on a computer such as a PC or workstation. The information processing program described in this embodiment is recorded on a computer-readable recording medium and executed by being read from the recording medium by the computer. The recording medium can be a hard disk, flexible disk, CD (Compact Disc)-ROM, MO (Magneto Optical Disc), DVD (Digital Versatile Disc), etc. Furthermore, the information processing program described in this embodiment may be distributed via a network such as the Internet.

[0143] 100 Information Processing Device 110 First Language Model 111,910 First Information 120 Second Language Model 121,920 Second Information 130 Third Language Model 131,1010 Third Information 140,500 Persona Information 141 Information 150,610 Question Information 160,1110,1120 Response Information 200 Information Processing System 201 Control Device 202 Language Processing Device 203 Client Device 210 Network 300 Bus 301 CPU 302 Memory 303 Network Interface 304 Recording Medium Interface 305 Recording Medium 306 Display 307 Input Device 400 Storage Unit 401 Acquisition Unit 402 Extraction Unit 403 First Generation Unit 404 Second Generation Unit 405 Third Generation Unit 406 Response generation unit 407 Output unit 600 LLM 620 Extracted persona information 700, 800 Template 701-706, 801-807 Location 1020 Fourth information 1030 Text 1040 Fifth information

Claims

1. An information processing program characterized by causing a computer to execute the following processes:

1. Obtaining question information related to a question; obtaining persona information including information for each of one or more items relating to the characteristics of the target; generating first information from the obtained question information that explains the content related to the obtained persona information; generating second information that explains the meaning of the obtained question information; generating third information based on the generated first information and the generated second information that explains the response to the question according to the information for each of the one or more items; and generating response information to the question information based on the generated third information.

2. The information processing program according to claim 1, characterized in that at least one of the processes among the process of generating the first information, the process of generating the second information, the process of generating the third information, and the process of generating the response information utilizes one or more language models that output text as a response to the input text.

3. The information processing program according to claim 2, wherein the process for acquiring the persona information is characterized in that, using the language model, based on the acquired question information, the information of one or more items relating to the characteristics that are related to the question is extracted from the information of each of the one or more items relating to the characteristics, thereby acquiring the persona information including the information of each of the extracted items relating to the characteristics.

4. The information processing program according to claim 3, wherein the process for generating the third information includes the generated first information and the generated second information, and when a prompt requesting the third information corresponding to the information of at least one of the one or more items is input to the language model, the program generates the third information corresponding to the information of at least one of the one or more items based on the text output by the language model, and repeats this process until the third information corresponding to the information of each of the one or more items is available.

5. The information processing program according to any one of 2 to 4, characterized in that the process for generating the response information uses the language model to generate the response information that summarizes and explains the generated third information according to the features.

6. The information processing program according to any one of claims 1 to 4, characterized in that the persona information includes profile or life story information relating to a specific person corresponding to the personality.

7. An information processing method characterized in that a computer performs the following processes: acquire question information relating to a question; acquire persona information including information for each of one or more items relating to the characteristics of the target; generate first information from the acquired question information that explains the content related to the acquired persona information; generate second information that explains the meaning of the acquired question information; generate third information based on the generated first information and the generated second information that explains the response to the question according to the information for each of the one or more items; and generate response information to the question information based on the generated third information.

8. An information processing device comprising a control unit that acquires question information relating to a question, acquires persona information including information for each of one or more items relating to the characteristics of the target, generates first information that explains the content of the acquired persona information from the acquired question information, generates second information that explains the meaning of the acquired question information, generates third information that explains the response to the question according to the information for each of the one or more items based on the generated first information and the generated second information, and generates response information to the question information based on the generated third information.

Images