Generation program, generation method, and information processing device

Abstract

This information processing device receives an input from a user. The information processing device uses an attribute of the user and a predetermined language model to generate, from the received input, a three-dimensional-space object. The information processing device generates a three-dimensional space on the basis of the generated three-dimensional-space object.

Description

GENERATION PROGRAM, GENERATION METHOD, AND INFORMATION PROCESSING APPARATUS

[0001] The present invention relates to a generation program and the like.

[0002] A variety of services are being offered that utilize a three-dimensional virtual space called the Metaverse (hereafter referred to as the "3D space"). For example, Metaverse users can move their own avatars around in the three-dimensional space, go shopping with friends, or hold meetings with colleagues.

[0003] Furthermore, there is a technology called digital twin, which represents physical objects in the real world in three-dimensional space. If digital twins can be used to accurately represent real-world objects such as products, clothing, rooms, and buildings, users can enjoy a more realistic experience. To achieve this, engineers set detailed information about the objects and perform verification work to increase the accuracy of the three-dimensional space.

[0004] “Hitachi uses metaverse to train nuclear power plant maintenance and inspection workers,” Nikkei Shimbun, [Retrieved September 1, 2023], Internet <URL: https: / / www.nikkei.com / article / DGXZQOUC0526Q0V01C22A2000000 / >

[0005] However, the above-mentioned prior art techniques leave room for improvement in terms of generating a harmonious set of three-dimensional spaces.

[0006] For example, expressing a complete, harmonious three-dimensional space based on a digital twin requires specialized knowledge of three-dimensional space, and is not something that can be easily created by the average user.

[0007] In one aspect, the present invention aims to provide a generation program, a generation method, and an information processing device that can generate a set of harmonious three-dimensional spaces.

[0008] In the first proposal, a computer is caused to execute the following process: the computer receives input from a user; the computer generates an object in a three-dimensional space from the received input using the user's attributes and a predetermined language model; and the computer generates a three-dimensional space based on the object in the generated three-dimensional space.

[0009] A set of harmonious three-dimensional spaces can be generated.

[0010] FIG. 1 is a diagram illustrating an example of processing by an information processing device according to the first embodiment. FIG. 2 is a diagram illustrating an example of a data structure of a vector DB according to the first embodiment. FIG. 3 is a diagram illustrating an example of a data structure of an asset DB according to the first embodiment. FIG. 4 is a diagram illustrating an example of three-dimensional space information generated by a generation unit according to the first embodiment. FIG. 5 is a diagram illustrating an example of adjusted three-dimensional space information. FIG. 6 is a functional block diagram illustrating a configuration of an information processing device according to the first embodiment. FIG. 7 is a flowchart illustrating a processing procedure of the information processing device according to the first embodiment. FIG. 8 is a diagram illustrating an example of processing by an information processing device according to the second embodiment. FIG. 9 is a diagram illustrating an example of a data structure of a company knowledge DB according to the second embodiment. FIG. 10 is a diagram illustrating an example of prompt information generated by a post-processing unit according to the second embodiment. FIG. 11 is a diagram illustrating an example of answer information generated by an answer generation unit according to the second embodiment. FIG. 12 is a functional block diagram illustrating a configuration of an information processing device according to the second embodiment. FIG. 13 is a flowchart illustrating a processing procedure of the information processing device according to the second embodiment. FIG. 14 is a diagram illustrating an example of processing by an information processing device according to the third embodiment. FIG. 15 is a diagram illustrating an example of processing by a prompt adjustment unit according to the third embodiment. FIG. 16 is a diagram illustrating an example of prompt information generated by a post-processing unit according to the third embodiment. FIG. 17 is a functional block diagram illustrating a configuration of an information processing device according to the third embodiment. FIG. 18 is a flowchart (1) illustrating a processing procedure of the information processing device according to the third embodiment. FIG. 19 is a flowchart (2) illustrating a processing procedure of the information processing device according to the third embodiment. FIG. 20 is a diagram for explaining processing of an information processing device according to the fourth embodiment. FIG. 21 is a diagram (1) illustrating an example of an interaction between an interviewer and an interviewee. FIG. 22 is a diagram (2) illustrating an example of an interaction between an interviewer and an interviewee. FIG. 23 is a diagram illustrating an example of prompt information set in an interviewer Agent. FIG. 24 is a diagram illustrating an example of prompt information set in an interviewee Agent. FIG. 25 is a diagram illustrating an example of persona information. FIG. 26 is a diagram illustrating an example of a result of a depth interview. FIG. 27 is a functional block diagram illustrating a configuration of an information processing device according to the fourth embodiment.FIG. 28 is a flowchart showing the processing procedure of an information processing device according to Example 4. FIG. 29 is a diagram for explaining characteristics of a depth interview and a group interview. FIG. 30 is a diagram for explaining processing of an information processing device according to Example 5. FIG. 31 is a diagram showing the results of a group interview. FIG. 32 is a diagram (2) for explaining processing of an information processing device according to Example 5. FIG. 33 is a diagram showing an example of product knowledge. FIG. 34 is a diagram showing the results of a depth interview. FIG. 35 is a functional block diagram showing the configuration of an information processing device according to Example 5. FIG. 36 is a flowchart showing the processing procedure of an information processing device according to Example 5. FIG. 37 is a diagram showing an example of the hardware configuration of a computer that realizes similar functions to the information processing device of Example 1. FIG. 38 is a diagram showing an example of the hardware configuration of a computer that realizes similar functions to the information processing device of Example 2. FIG. 39 is a diagram showing an example of the hardware configuration of a computer that realizes similar functions to the information processing device of Example 3. FIG. 40 is a diagram showing an example of the hardware configuration of a computer that realizes similar functions to the information processing device of Example 4. FIG. 41 is a diagram showing an example of the hardware configuration of a computer that realizes similar functions to the information processing device of Example 5.

[0011] Hereinafter, embodiments of the generating program, generating method, and information processing device disclosed in the present application will be described in detail with reference to the accompanying drawings. However, the present invention is not limited to these embodiments.

[0012] An example of processing by an information processing device according to the first embodiment will be described. In the first embodiment, as an example, a case will be described in which the information processing device automatically generates a complete set of three-dimensional space of a room based on information input by a user. The information processing device according to the first embodiment will be referred to as an "information processing device 100."

[0013] 1 is a diagram illustrating an example of processing by an information processing device according to Example 1. As illustrated in Fig. 1, the information processing device 100 includes a display unit 130, a vector DB 141, and an asset DB 142. The information processing device 100 also includes an attribute extraction unit 151, an adjustment query extraction unit 152, a search unit 153, a generation unit 154, and an adjustment unit 155.

[0014] The information processing device 100 receives input of user information 10a, text information 10b, and floor plan information 20 from a user who operates the input unit. The text information 10c will be described later.

[0015] The user information 10a is text describing the attributes of the user. The user attributes are information indicating the characteristics of the user, and include, for example, the user's age, gender, address, occupation, hobbies, department, job title, etc. The information processing device 100 may associate the user information 10a with a user ID, etc., so that the user can acquire the user information 10a corresponding to the user ID by specifying the user ID.

[0016] The text information 10b is text describing what kind of room the user wants to create. For example, the text information 10b may be an ambiguous expression such as "I want to create a room that gives off an ocean vibe." The user may also input the content of the text information 10b to the information processing device 100 by voice information. When the information processing device 100 acquires the voice information, it converts the voice information into text information 10b using voice recognition technology.

[0017] The floor plan information 20 is information that specifies the layout of a room, such as a living room, a bedroom, etc.

[0018] The vector DB 141 is a database that stores vectors corresponding to each attribute. The vectors corresponding to each attribute are assumed to be set in advance. FIG. 2 is a diagram illustrating an example of the data structure of the vector DB according to the first embodiment. As shown in FIG. 2, the vector DB 141 associates attributes with vectors.

[0019] The asset DB 142 is a database that stores information about multiple three-dimensional assets. Three-dimensional assets are furniture such as sofas, tables, chairs, desks, beds, cabinets, shelves, and rugs, as well as furniture or home appliances such as televisions, radios, and lights. Three-dimensional assets may also include doors, windows, wallpaper, and the like. In the following description, three-dimensional assets are simply referred to as "assets."

[0020] FIG. 3 is a diagram illustrating an example of a data structure of the asset DB according to the first embodiment. As illustrated in FIG. 3, the asset DB 142 associates indexes with asset information. The asset DB 142 uses vectors as indexes. The asset information includes, as asset information, initial values ​​of asset parameters, asset identification information, and the like. The asset parameters include the shape, size, color, and the like of the asset.

[0021] When the attribute extraction unit 151 receives input of the user information 10a, it analyzes the user information 10a to extract attributes contained in the user information 10a. The attribute extraction unit 151 outputs information on the extracted attributes to the search unit 153. In the following description, the attribute information output by the attribute extraction unit 151 to the search unit 153 will be referred to as "attribute information." The attribute extraction unit 151 is, for example, a pre-trained LLM (Large Language Model).

[0022] Upon receiving the input of the text information 10b, the adjusted query extraction unit 152 analyzes the text information 10b to extract the user's preferences. The adjusted query extraction unit 152 outputs the extracted preference information to the search unit 153. In the following description, the preference information output by the adjusted query extraction unit 152 to the search unit 153 will be referred to as "preference information." The adjusted query extraction unit 152 is an LLM or the like that is trained in advance using text information as input and preference information as output.

[0023] For example, if the text information 10b is "I want a room that gives off an ocean vibe," the adjustment query extraction unit 152 extracts preference information "wallpaper color=blue" and outputs it to the search unit 153. For example, blue is the color of the ocean. Also, if the text information 10b is "I want a bigger bed in my room," the adjustment query extraction unit 152 extracts preference information "bed size=sizeC" and outputs it to the search unit 153. Note that multiple sizes of assets such as beds are defined in advance, and the size relationship is, for example, "sizeA<sizeB<sizeC<...".

[0024] The search unit 153 searches for multiple pieces of asset information constituting a complete three-dimensional space based on the attribute information and the preference information. For example, the search unit 153 identifies a vector corresponding to an attribute based on the attribute included in the attribute information and the vector DB. The search unit 153 compares the identified attribute vector with an index (vector) in the asset DB 142 to search for asset information corresponding to the attribute vector. When the attribute information includes multiple attributes, the search unit 153 searches for multiple pieces of asset information by repeatedly executing the above process. Note that the search unit 153 may calculate the cosine similarity between the vector corresponding to the attribute and the index (vector) in the asset DB 142, and search for asset information corresponding to an index whose cosine similarity is equal to or greater than a threshold.

[0025] Furthermore, the search unit 153 updates the parameters of the searched asset information based on the preference information. For example, if preference information "wallpaper color=blue" is acquired and the searched asset information includes a wallpaper asset, the search unit 153 updates the parameters of the asset information. For example, the search unit 153 updates the wallpaper color (initial value) included in the parameters of the asset information to blue. If the search unit 153 has acquired multiple pieces of preference information, it updates the parameters of each piece of asset information by repeatedly executing the above process.

[0026] The search unit 153 outputs the asset information searched for based on the attribute information and having parameters updated based on the preference information to the generation unit 154 as a search result.

[0027] The generation unit 154 generates a set of three-dimensional spaces based on the floor plan information 20 and the search results of the search unit 153. In the following description, the information on the set of three-dimensional spaces is referred to as "three-dimensional space information." For example, in the floor plan information 20, the installation position of each asset is defined in advance, and the generation unit 154 generates three-dimensional space information by arranging assets included in the search results at the predefined asset installation positions. When arranging an asset, the generation unit 154 sets the shape, size, and color of the asset based on the asset parameters.

[0028] Fig. 4 is a diagram illustrating an example of three-dimensional space information generated by the generation unit according to the present embodiment 1. In the example illustrated in Fig. 4, assets 40a, 40b, 40c, 40d, 40e, 40f, 40g, and 40h included in the search results are set in three-dimensional space information 40.

[0029] The generation unit 154 outputs the generated three-dimensional space information to the adjustment unit 156 .

[0030] The adjustment unit 155 outputs the three-dimensional space information acquired from the generation unit 154 to the display unit 130 for display. The user refers to the three-dimensional space information displayed on the display unit 130, and if there is an adjusted portion, outputs the text information 10c to the adjusted query extraction unit 152 of the information processing device 100.

[0031] For example, the text information 10c is "I want to make the TV bigger."

[0032] When the adjustment query extraction unit 152 receives the input of the text information 10c, it analyzes the text information 10c to extract the user's preferences (preference information). The adjustment query extraction unit 152 generates adjustment parameters based on the extracted preference information and outputs them to the adjustment unit 155. For example, when the adjustment query extraction unit 152 generates the preference information "TV size=sizeD", it sets information to the adjustment parameter indicating "adjust the TV size to sizeD".

[0033] When the adjustment unit 155 acquires the adjustment parameters, it adjusts the three-dimensional space information based on the adjustment parameters and displays the adjusted three-dimensional space information on the display unit 130. The adjusted query extraction unit 152 and the adjustment unit 155 repeatedly execute the above process every time they acquire text information 10c from the user.

[0034] When a user requests the information processing device 100 to add a new asset, the information processing device 100 may add the requested asset to the three-dimensional space information. For example, when a new asset is added to the preference information extracted from the text information 10c, the adjustment query extraction unit 152 obtains information about the corresponding asset from the asset DB 142 and registers it in the adjustment parameters.

[0035] 5 is a diagram showing an example of three-dimensional space information before and after adjustment. Comparing three-dimensional space information 40 with three-dimensional space information 41, asset 40c has been removed from three-dimensional space information 41, and assets 41a and 41b have been added.

[0036] As described above, when the information processing device 100 according to the first embodiment receives user information 10a and text information 10b from a user, the information processing device 100 uses a predetermined language model to extract attribute information and preference information from the user information 10a and text information 10b. The information processing device 100 generates objects (assets) in a three-dimensional space based on the attribute information, updates the assets based on the preference information, and generates three-dimensional space information based on the updated objects. This makes it possible to easily generate a complete three-dimensional space even if the user provides vague instructions using the text information 10b.

[0037] Furthermore, when the information processing device 100 receives input of text information 10c from a user after displaying the three-dimensional space information, the information processing device 100 generates adjustment parameters based on the text information 10c and adjusts the three-dimensional space information, thereby making it possible to make the three-dimensional space information closer to the information preferred by the user.

[0038] Next, a configuration example of the information processing device 100 that executes the above-described processing will be described. Fig. 6 is a functional block diagram showing the configuration of the information processing device according to the first embodiment. As shown in Fig. 6, the information processing device 100 includes a communication unit 110, an input unit 120, a display unit 130, a storage unit 140, and a control unit 150.

[0039] The communication unit 110 executes data communication with external devices via a network, and is realized by a network interface card (NIC) or the like.

[0040] The input unit 120 is an input device that inputs various types of information to the information processing device 100. The input unit 120 corresponds to a keyboard, a mouse, a touch panel, etc. For example, a user operates the input unit 120 to input user information 10a, text information 10b and 10c, floor plan information 20, etc. to the control unit 150.

[0041] The display unit 130 is a display device that displays information output from the control unit 150. The display unit 130 corresponds to a head-mounted display, a liquid crystal display, an organic EL (Electro Luminescence) display, a touch panel, etc. For example, the display unit 130 displays three-dimensional space information 40, 41, etc., shown in Figures 4, 5, etc.

[0042] The storage unit 140 includes a vector DB 141 and an asset DB 142. The storage unit 140 is realized by, for example, a semiconductor memory element such as a random access memory (RAM) or a flash memory, or a storage device such as a hard disk or an optical disk.

[0043] The vector DB 141 is a database that stores vectors corresponding to each attribute. The description of the vector DB 141 is the same as the description of the vector DB 141 described with reference to FIG.

[0044] The asset DB 142 is a database that stores information about a plurality of three-dimensional assets. The description of the asset DB 142 is the same as the description of the asset DB 142 described with reference to FIG.

[0045] The control unit 150 includes an attribute extraction unit 151, an adjusted query extraction unit 152, a search unit 153, a generation unit 154, and an adjustment unit 155. The control unit 150 is realized by, for example, a central processing unit (CPU) or a micro processing unit (MPU). The control unit 150 may also be implemented by an integrated circuit such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA).

[0046] When the attribute extraction unit 151 receives input of user information 10a from the input unit 120, it analyzes the user information 10a to extract attributes contained in the user information 10a. The attribute extraction unit 151 outputs the attribute information to the search unit 153. Other details regarding the attribute extraction unit 151 are the same as those described in FIG. 1 .

[0047] When the adjusted query extraction unit 152 receives the input of the text information 10b from the input unit 120, the adjusted query extraction unit 152 analyzes the text information 10b to extract the user's preferences. The adjusted query extraction unit 152 outputs the preference information to the search unit 153.

[0048] After the three-dimensional space information is displayed on the display unit 130, the adjusted query extraction unit 152 receives input of text information 10c from the input unit 120 and analyzes the text information 10c to extract the user's preferences. The adjusted query extraction unit 152 generates adjustment parameters based on the preference information and outputs the adjustment parameters to the adjustment unit 155.

[0049] Other details regarding the adjusted query extraction unit 152 are the same as those regarding the adjusted query extraction unit 152 described with reference to FIG.

[0050] The search unit 153 searches for a plurality of pieces of asset information that make up a set of three-dimensional space based on the attribute information and the preference information, and outputs the search results to the generation unit 154. Other explanations regarding the search unit 153 are the same as those regarding the search unit 153 described in FIG. 1 .

[0051] The generation unit 154 acquires floor plan information 20 via the input unit 120. The generation unit 154 also acquires search results from the search unit 153. The generation unit 154 generates three-dimensional space information based on the floor plan information 20 and the search results of the search unit 153. The generation unit 154 outputs the three-dimensional space information to the adjustment unit 155. Other explanations regarding the generation unit 154 are the same as those regarding the generation unit 154 described in FIG. 1 .

[0052] The adjustment unit 155 outputs the three-dimensional spatial information acquired from the generation unit 154 to the display unit 130 for display. Furthermore, when the adjustment unit 155 acquires adjustment parameters from the adjusted query extraction unit 152, the adjustment unit 155 adjusts the three-dimensional spatial information using the adjustment parameters, and outputs the adjusted three-dimensional spatial information to the display unit 130 for display. Other explanations regarding the adjustment unit 155 are the same as those regarding the adjustment unit 155 described in FIG. 1 .

[0053] Next, an example of a processing procedure of the information processing device 100 according to the present embodiment will be described. Fig. 7 is a flowchart showing the processing procedure of the information processing device according to the present embodiment. As shown in Fig. 7, the information processing device 100 receives input of user information 10a, text information 10b, and floor plan information 20 (step S101).

[0054] The attribute extraction unit 151 of the information processing device 100 extracts attribute information based on the user information 10a (step S102). The adjusted query extraction unit 152 of the information processing device 100 extracts preference information based on the text information (step S103).

[0055] The search unit 153 of the information processing device 100 identifies a vector corresponding to the attribute based on the attribute information and the vector DB 141 (step S104). The search unit 153 searches the asset DB 142 for asset information based on the vector corresponding to the attribute (step S105).

[0056] The search unit 153 updates the parameters of the assets included in the asset information based on the preference information (step S106). The generation unit 154 of the information processing device 100 generates three-dimensional space information based on the search results of the search unit 153 and the floor plan information 20 (step S107).

[0057] The adjustment unit 155 of the information processing device 100 displays the three-dimensional space information on the display unit 130 (step S108). When the adjusted query extraction unit 152 receives new text information 10c (step S109, Yes), the adjusted query extraction unit 152 extracts preference information based on the text information 10c (step S110).

[0058] The adjustment query extraction unit 152 generates adjustment parameters based on the preference information (step S111). The adjustment unit 155 updates the three-dimensional space information based on the adjustment parameters (step S112), and the process proceeds to step S108.

[0059] On the other hand, if the information processing device 100 has not received new text information 10c (No at step S109), the process ends.

[0060] The effects of the information processing device 100 according to the first embodiment will be described. When the information processing device 100 receives user information 10a and text information 10b from a user, the information processing device 100 uses a predetermined language model to extract attribute information and preference information from the user information 10a and text information 10b. The information processing device 100 generates objects (assets) in a three-dimensional space based on the attribute information, updates the assets based on the preference information, and generates three-dimensional space information based on the updated objects. This makes it possible to easily generate a complete three-dimensional space even if the user gives vague instructions using the text information 10b.

[0061] Furthermore, when the information processing device 100 receives input of text information 10c from a user after displaying the three-dimensional space information, the information processing device 100 generates adjustment parameters based on the text information 10c and adjusts the three-dimensional space information, thereby making it possible to make the three-dimensional space information closer to the information preferred by the user.

[0062] Next, an example of processing by an information processing device according to the second embodiment will be described. In the second embodiment, as an example, the information processing device outputs a response to a user based on information (question) input by the user. The information processing device according to the second embodiment will be referred to as "information processing device 200."

[0063] 8 is a diagram illustrating an example of processing by the information processing device according to Example 2. As shown in FIG. 8 , the information processing device 200 includes a display unit 230, a company knowledge DB 241, a role identification unit 251, a preprocessing unit 252, a search unit 253, a postprocessing unit 254, a response generation unit 255, and an image generation unit 256.

[0064] The information processing device 200 receives input of text information 15 and user information 16 from a user who operates the input unit.

[0065] The text information 15 is text information in which a question from the user is written. For example, the text information 15 may be, "Recently, I've heard that the company is changing. How does this relate to my work?"

[0066] The user information 16 is text that describes the user's attributes, business details, etc. The user's attributes are information that indicates the user's characteristics, and include, for example, the user's age, gender, address, occupation, hobbies, department, job title, etc. In the second embodiment, information that compiles the user's attributes and business details is referred to as the user's "role."

[0067] The company knowledge DB 241 is a database that stores business knowledge information related to the user's company. Fig. 9 is a diagram showing an example of the data structure of the company knowledge DB according to the second embodiment. As shown in Fig. 9, the company knowledge DB 241 associates indexes with business knowledge information. The company knowledge DB 241 uses vectors as indexes.

[0068] In the individual company knowledge DB 241, the business knowledge information set includes business data related to the user's business, text information of past statements made by the user's superior or a person equivalent to the management team of the user's company, text information of current issues related to the department to which the user belongs, text information of the mid-term plan of the department to which the user belongs, etc. Furthermore, the business knowledge information may also include the above-described user attributes and business content, and text information from when the information processing device 200 was used in the past.

[0069] Returning to the description of FIG. 8 , the role identification unit 251 identifies a role when receiving input of user information 16. For example, the role identification unit 251 identifies a role corresponding to the user information 16 using an identification table that defines the relationship between user attributes and roles. Note that the role identification unit 251 may identify a role corresponding to the user information 16 using a trained neural network (NN) that inputs user attributes and outputs roles. Alternatively, the role may be determined by an in-house (private) LLM using the history of text information 15 from the user. In this second embodiment, the role is represented by text.

[0070] The role specifying unit 251 outputs the role information to the search unit 253. The role specified by the role specifying unit 251 from the user information 16 is set in the role information.

[0071] When the preprocessing unit 252 receives the text information 15, it converts the text information 15 into a predetermined prompt format. For example, the text information 15 may be "Recently, people have been saying that they are changing companies. How does this relate to my work?" In this case, the preprocessing unit 252 converts the text information 15 into a predetermined prompt format such as "The user's question is as follows: Recently, people have been saying that they are changing companies. How does this relate to my work?"

[0072] The preprocessing unit 252 outputs the converted prompt format information to the search unit 253 as a preprocessing result.

[0073] The search unit 253 acquires role information, and compares a vector index generated by combining the preprocessing result of the preprocessing unit 252 with the role information and inputting the combined result to the embeddings processing unit 253a with the index of the individual company knowledge DB 241 to search for business knowledge information related to the text information input by the user. For example, the search unit 253 identifies, among the vectors set as indexes, an index that has the highest cosine similarity with the text vector combining the preprocessing result with the role information, and searches for business knowledge information corresponding to the identified index. The search unit 253 outputs the searched business knowledge information, the preprocessing result, and the role information to the postprocessing unit 254 as search results.

[0074] The embeddings processing unit 253a performs a process of vectorizing the text information input from the search unit 253. For example, the embeddings processing unit 253a combines the preprocessing result with role information to vectorize the information.

[0075] When the post-processing unit 254 acquires the search results (input text information + business knowledge related to the role), it generates prompt information to be input to the private LLM and to obtain a response from the private LLM. Fig. 10 is a diagram showing an example of prompt information generated by the post-processing unit according to the second embodiment. As shown in Fig. 10, the prompt information 30 generated by the post-processing unit 254 includes prompts 31, 32, 33, 34, and 35.

[0076] Prompt 31 is text information corresponding to the content of the above preprocessing result. Prompts 32 to 35 are business knowledge information corresponding to the user searched for by the search unit 253. In addition to the information shown in FIG. 10 , the post-processing unit 254 may set a supplementary sentence such as "answer the question in prompt 31 based on prompts 32 to 35" in the prompt information 30.

[0077] The post-processing unit 254 outputs the generated prompt (for example, prompt information 30 ) to the answer generating unit 255 .

[0078] The answer generation unit 255 is a processing unit that generates answer information to a prompt acquired from the post-processing unit 254. For example, the answer generation unit 255 further uses the contents of prompts 32 to 35 when generating answer information to prompt 31 included in prompt information 30. For example, the answer generation unit 255 is a private LLM. The answer generation unit 255 outputs the answer information to the image generation unit 256.

[0079] Fig. 11 is a diagram illustrating an example of answer information generated by the answer generation unit according to Example 2. In the example illustrated in Fig. 11, phrases 61, 62, 63, and 64 included in answer information 60 are difficult to generate directly from prompt 31 included in prompt information 30, but can be added by using prompts 32 to 35.

[0080] The image generation unit 256 generates a 3D face model using a pre-specified face image sample or the like, and outputs the generated 3D face model to the display unit 230 for display. Furthermore, the image generation unit 256 converts the answer information generated by the answer generation unit 255 into audio information using a pre-specified audio sample or the like, and outputs the audio information from a speaker or the like. Note that the image generation unit 256 may output the answer information to the display unit 230 for display.

[0081] As described above, when the information processing device 200 according to the second embodiment receives the text information 15 and the user information 16 from a user, it identifies the role corresponding to the user information 16 and searches for business knowledge information from the individual company knowledge DB 241. Based on the text information 15 and the business knowledge information of the user, the information processing device 200 generates prompt information that allows an answer to be obtained from the answer generation unit 255 (private LLM). This makes it possible to generate prompt information that allows the user who has asked a question to receive a convincing answer.

[0082] Next, a description will be given of a configuration example of an information processing device 200 according to this embodiment 2. Fig. 12 is a functional block diagram showing the configuration of the information processing device according to this embodiment 2. As shown in Fig. 12, the information processing device 200 includes a communication unit 210, an input unit 220, a display unit 230, a speaker 235, a storage unit 240, and a control unit 250.

[0083] The communication unit 210 executes data communication with external devices via a network, and is realized by a NIC or the like.

[0084] The input unit 220 is an input device that inputs various types of information to the information processing device 200. The input unit 220 corresponds to a keyboard, a mouse, a touch panel, etc. For example, a user operates the input unit 220 to input text information 15, user information 16, etc. to the control unit 250.

[0085] The display unit 230 is a display device that displays information received from the control unit 250. The display unit 230 corresponds to a head-mounted display, a liquid crystal display, an organic EL display, a touch panel, etc. For example, the display unit 230 displays the 3D face model generated by the image generation unit 256, answer information, etc.

[0086] The speaker 235 outputs audio information received from the control unit 250 .

[0087] The storage unit 240 includes a company knowledge DB 241. The storage unit 240 is realized by, for example, a semiconductor memory element such as a RAM or a flash memory, or a storage device such as a hard disk or an optical disk.

[0088] The individual company knowledge DB 241 is a database that holds business knowledge information related to the user's company, etc. The explanation regarding the individual company knowledge DB 241 is the same as that explained with reference to FIG.

[0089] The control unit 250 includes a role identification unit 251, a preprocessing unit 252, a search unit 253, a postprocessing unit 254, a response generation unit 255, and an image generation unit 256. The control unit 250 is realized by, for example, a CPU or an MPU. The control unit 250 may also be implemented by, for example, an integrated circuit such as an ASIC or an FPGA.

[0090] The role identification unit 251 identifies a role when receiving input of the user information 16. The role identification unit 251 outputs the role information to the search unit 253. Other details regarding the role identification unit 251 are the same as those regarding the role identification unit 251 described in FIG. 8 .

[0091] When the preprocessing unit 252 receives the text information 15, it converts the text information 15 into a preset prompt format and generates a preprocessing result. The preprocessing unit 252 outputs the preprocessing result to the search unit 253. Other details regarding the preprocessing unit 252 are the same as those described in FIG. 8 .

[0092] The search unit 253 acquires role information and searches for business knowledge information by comparing the role (vector) set in the acquired role information with the index of the individual company knowledge DB 241. The search unit 253 outputs the searched business knowledge information and the preprocessing result to the post-processing unit 254 as a search result. Other explanations regarding the search unit 253 are the same as those regarding the search unit 253 described in FIG. 8. Although not shown in the figure, the search unit 253 is connected to the embeddings processing unit 253a described in FIG. 8. Note that the search unit 253 may have the function of the embeddings processing unit 253a.

[0093] When the post-processing unit 254 acquires the search results (business knowledge information, pre-processing results), it generates prompt information to be input to the private LLM and to obtain an answer from the private LLM. The post-processing unit 254 outputs the prompt information to the answer generation unit 255. Other details regarding the post-processing unit 254 are the same as those described in FIG. 8 .

[0094] The answer generation unit 255 is a processing unit that generates answer information to a prompt acquired from the post-processing unit 254. For example, the answer generation unit 255 is a private LLM. The answer generation unit 255 outputs the answer information to the image generation unit 256. Other details regarding the answer generation unit 255 are the same as those described in FIG. 8 .

[0095] The image generation unit 256 generates a 3D face model and outputs the generated 3D face model to the display unit 230 for display. The image generation unit 256 also converts the answer information generated by the answer generation unit 255 into audio information and outputs it from the speaker 235. The image generation unit 256 may also output the answer information to the display unit 230 for display. Other details regarding the image generation unit 256 are the same as those regarding the image generation unit 256 described in FIG. 8.

[0096] Next, an example of a processing procedure of the information processing device 200 according to the present embodiment 2 will be described. Fig. 13 is a flowchart showing the processing procedure of the information processing device according to the present embodiment 2. As shown in Fig. 13, the role identification unit 251 of the information processing device 200 receives input of user information 16 and identifies a role (step S201).

[0097] The preprocessing unit 252 of the information processing device 200 receives the text information 15 and converts it into a preset prompt format (step S202). The search unit 253 of the information processing device 200 compares the role information with the individual company knowledge DB 241 to search for business knowledge information (step S203).

[0098] The post-processing unit 254 of the information processing device 200 generates prompt information based on the searched business knowledge information and the pre-processing result (step S204). The answer generation unit 255 of the information processing device 200 generates answer information based on the prompt information (step S205).

[0099] The image generating unit 256 of the information processing device 200 converts the answer information into voice information (step S206). The image generating unit 256 displays the 3D face model on the display unit 230 and outputs the voice information from the speaker 235 (step S207).

[0100] If new text information has been received (step S208, Yes), the information processing device 200 proceeds to step S202. On the other hand, if new text information has not been received (step S208, No), the information processing device 200 ends the process.

[0101] The effects of the information processing device 200 according to the second embodiment will be described. When the information processing device 200 receives text information 15 and user information 16 from a user, it identifies the role corresponding to the user information 16 and searches for business knowledge information from the company knowledge DB 241. Based on the user's text information 15 and business knowledge information, the information processing device 200 generates prompt information that allows an answer to be obtained from the answer generation unit 255 (private LLM). This makes it possible to generate prompt information that allows a user who has asked a question to receive a convincing answer.

[0102] Next, an example of processing by an information processing device according to the third embodiment will be described. The information processing device according to the third embodiment will be referred to as an "information processing device 300." The information processing device 300 has a function that combines the functions of the information processing device 100 and the information processing device 200 described above.

[0103] 14 is a diagram illustrating an example of processing by an information processing device according to Example 3. As illustrated in Fig. 14 , the information processing device 300 includes a display unit 330, a vector DB 341, and an asset DB 342. The information processing device 300 also includes an attribute extraction unit 351, an adjustment query extraction unit 352, a search unit 353, a generation unit 354, an adjustment unit 355, and a prompt adjustment unit 400.

[0104] The information processing device 300 accepts input of user information 10a, text information 10b, and floor plan information 20 from a user operating the input unit. After displaying three-dimensional space information on the display unit 330, the information processing device 300 accepts input of text information 10c from a user operating the input unit. Other explanations regarding the user information 10a, text information 10b and 10c, and floor plan information 20 are the same as those regarding the user information 10a, text information 10b and 10c, and floor plan information 20 described in FIG. 1 .

[0105] The prompt adjustment unit 400 receives the user information 10a and the text information 10b and generates an adjustment prompt. For example, the prompt adjustment unit 400 identifies a role corresponding to the user information 10a and adds business knowledge information corresponding to the role to the text information 10b to generate an adjustment prompt. The prompt adjustment unit 400 outputs the adjustment prompt to the adjustment query extraction unit 352.

[0106] 15 is a diagram illustrating an example of processing by the prompt adjustment unit according to the present embodiment 3. As illustrated in FIG. 15 , the prompt adjustment unit 400 includes a company knowledge DB 441, a role identification unit 451, a preprocessing unit 452, a search unit 453, and a postprocessing unit 454.

[0107] The individual company knowledge DB 441 is a database that holds business knowledge information related to the user's company. The individual company knowledge DB 441 associates indexes with business knowledge information in the same manner as the individual company knowledge DB 241 described in Fig. 9. The individual company knowledge DB 441 uses vectors as indexes.

[0108] In the individual company knowledge DB 441 according to the third embodiment, text information relating to the work environment when an employee telecommutes, which has been previously stated by a person equivalent to the superior of a user corresponding to a role, is set as business knowledge information. The text information on the work environment is the color, shape, size, etc. of a specific asset. For example, the text information on the work environment is, "It is better to use a larger desk."

[0109] When the role identification unit 451 receives input of the user information 10a, it identifies a role corresponding to the user information 10a and outputs the role information to the search unit 453. Other details regarding the role identification unit 451 are the same as those regarding the role identification unit 251 described in FIG. 8 .

[0110] When the preprocessing unit 452 receives the text information 10b, it converts the text information 10b into a predetermined prompt format. For example, if the text information 10b is "I want a room that gives off an ocean vibe," the preprocessing unit 452 converts the text information 10b into a predetermined prompt format such as "The user's request is as follows: I want the wallpaper in my room to be blue."

[0111] The preprocessing unit 452 outputs the converted prompt format information to the search unit 453 as a preprocessing result.

[0112] The search unit 453 acquires role information and searches for business knowledge information by comparing the role (vector) set in the acquired role information with the index of the individual company knowledge DB 441. For example, the search unit 453 identifies an index that has the largest cosine similarity with the role vector among the vectors set as indexes, and searches for business knowledge information corresponding to the identified index. The search unit 453 outputs the searched business knowledge information and the preprocessing result to the postprocessing unit 454 as the search result.

[0113] Upon acquiring the search results (business knowledge information, preprocessing results), the post-processing unit 454 generates an adjustment prompt to be input to the private LLM and to obtain an answer (preference information) from the private LLM. FIG. 16 is a diagram illustrating an example of an adjustment prompt generated by the post-processing unit according to the third embodiment. As shown in FIG. 16, the adjustment prompt 50 generated by the post-processing unit 454 includes prompts 51 and 52.

[0114] The prompt 51 is text information corresponding to the content of the preprocessing result, and the prompt 52 is business knowledge information corresponding to the user searched by the search unit 453.

[0115] The post-processing unit 454 outputs the generated adjustment prompt 50 to the adjustment query extraction unit 352 in FIG.

[0116] Here, when the prompt adjustment unit 400 receives text information 10c after the three-dimensional spatial information is displayed on the display unit 330, it generates an adjustment prompt in the same manner as in the case of the text information 10b described above. If the user is the same, the prompt adjustment unit 400 reuses the previously acquired user information 10a. The prompt adjustment unit 400 outputs the generated adjustment prompt to the adjustment query extraction unit 352. The prompt adjustment unit 400 repeats the above process when receiving other text information.

[0117] Returning to the description of Fig. 14, when the attribute extraction unit 351 receives input of the user information 10a, it analyzes the user information 10a to extract attributes contained in the user information 10a. The attribute extraction unit 351 outputs the attribute information to the search unit 353. Other details regarding the attribute extraction unit 351 are the same as those regarding the attribute extraction unit 151 described in Fig. 1.

[0118] When the adjusted query extraction unit 352 receives an input of an adjustment prompt (adjustment prompt 50), it analyzes the adjustment prompt to extract the user's preferences. The adjusted query extraction unit 352 outputs information on the user's preferences as "preference information" to the search unit 353. The adjusted query extraction unit 152 is an LLM or the like that is trained in advance, with text information as input and preference information as output.

[0119] For example, when the adjustment prompt includes prompt 51, the adjustment query extraction unit 352 extracts the preference information “wallpaper color=blue” and outputs it to the search unit 353. When the adjustment prompt includes prompt 52, the adjustment query extraction unit 352 extracts the preference information “desk size=sizeC” and outputs it to the search unit 353.

[0120] Furthermore, when the adjustment query extraction unit 352 receives an adjustment prompt from the prompt adjustment unit 400 after the three-dimensional space information is displayed on the display unit 330, it extracts user preference information in the same manner as the above-described process. The adjustment query extraction unit 352 generates adjustment parameters based on the extracted preference information and outputs them to the adjustment unit 355. The process by which the adjustment query extraction unit 352 generates adjustment parameters is the same as that of the adjustment query extraction unit 152 described in the first embodiment.

[0121] The search unit 353 searches for a plurality of pieces of asset information that make up a set of three-dimensional space based on the attribute information and preference information, and outputs the search results to the generation unit 354. Other explanations regarding the search unit 353 are the same as those regarding the search unit 153 described in FIG. 1.

[0122] The generation unit 354 acquires the floor plan information 20. The generation unit 354 also acquires search results from the search unit 353. The generation unit 354 generates three-dimensional space information based on the floor plan information 20 and the search results of the search unit 353. The generation unit 354 outputs the three-dimensional space information to the adjustment unit 355. Other explanations regarding the generation unit 354 are the same as those regarding the generation unit 154 described in FIG. 1 .

[0123] The adjustment unit 355 outputs the three-dimensional spatial information acquired from the generation unit 354 to the display unit 330 for display. Furthermore, when the adjustment unit 355 acquires adjustment parameters from the adjusted query extraction unit 352, the adjustment unit 355 adjusts the three-dimensional spatial information using the adjustment parameters, and outputs the adjusted three-dimensional spatial information to the display unit 330 for display. Other explanations regarding the adjustment unit 355 are the same as those regarding the adjustment unit 155 described in FIG. 1 .

[0124] As described above, the information processing device 300 according to the third embodiment receives input from a user (user information 10a, text information 10b, etc.), searches for assets in a three-dimensional space using the user's attributes and a predetermined language model, and generates three-dimensional space information. This makes it possible to easily generate a complete three-dimensional space even if the user inputs vague instructions.

[0125] Next, a configuration example of an information processing device 300 that executes the above-described processing will be described. Fig. 17 is a functional block diagram showing the configuration of an information processing device according to Example 3. As shown in Fig. 17, the information processing device 300 includes a communication unit 310, an input unit 320, a display unit 330, a storage unit 340, and a control unit 350.

[0126] The communication unit 310 executes data communication with external devices via a network, and is realized by a NIC or the like.

[0127] The input unit 320 is an input device that inputs various types of information to the information processing device 300. The input unit 320 corresponds to a keyboard, a mouse, a touch panel, etc. For example, a user operates the input unit 320 to input user information 10a, text information 10b and 10c, floor plan information 20, etc. to the control unit 350.

[0128] The display unit 330 is a display device that displays information output from the control unit 350. The display unit 330 corresponds to a head-mounted display, a liquid crystal display, an organic EL display, a touch panel, etc. For example, the display unit 330 displays three-dimensional space information output from the adjustment unit 355, etc.

[0129] The storage unit 340 includes a vector DB 341, an asset DB 342, and an individual company knowledge DB 441. The storage unit 340 is realized by, for example, a semiconductor memory element such as a RAM or a flash memory, or a storage device such as a hard disk or an optical disk.

[0130] The vector DB 341 is a database that stores vectors corresponding to each attribute. The description of the vector DB 341 is the same as the description of the vector DB 141 described with reference to FIG. 2 in the first embodiment.

[0131] The asset DB 342 is a database that stores information about a plurality of three-dimensional assets. The description of the asset DB 342 is the same as the description of the asset DB 142 described in FIG.

[0132] The individual company knowledge DB 441 is a database that stores business knowledge information related to the user's company, etc. Other details regarding the individual company knowledge DB 441 are the same as those described above for the individual company knowledge DB 441 in the third embodiment.

[0133] The control unit 350 includes an attribute extraction unit 351, an adjusted query extraction unit 352, a search unit 353, a generation unit 354, an adjustment unit 355, and a prompt adjustment unit 400. The control unit 350 is realized by, for example, a CPU or an MPU. Alternatively, the control unit 350 may be implemented by, for example, an integrated circuit such as an ASIC or an FPGA.

[0134] When the attribute extraction unit 351 receives input of user information 10a from the input unit 320, it analyzes the user information 10a to extract attributes contained in the user information 10a. The attribute extraction unit 351 outputs the attribute information to the search unit 353. Other details regarding the attribute extraction unit 351 are the same as those described in FIG. 14 .

[0135] Upon receiving the input of the adjustment prompt (adjustment prompt 50), the adjustment query extraction unit 352 analyzes the adjustment prompt to extract the user's preferences. The adjustment query extraction unit 352 outputs the preference information to the search unit 353.

[0136] When the adjustment query extraction unit 352 receives an adjustment prompt (adjustment prompt 50) after the three-dimensional space information is displayed on the display unit 330, the adjustment query extraction unit 352 analyzes the text information 10c to extract the user's preferences. The adjustment query extraction unit 352 generates adjustment parameters based on the preference information and outputs the adjustment parameters to the adjustment unit 355.

[0137] Other details regarding the adjusted query extraction unit 352 are the same as those regarding the adjusted query extraction unit 352 described with reference to FIG.

[0138] The search unit 353 searches for a plurality of pieces of asset information that make up a set of three-dimensional space based on the attribute information and the preference information, and outputs the search results to the generation unit 354. Other explanations regarding the search unit 353 are the same as those regarding the search unit 353 described in FIG. 14 .

[0139] The generation unit 354 acquires the floor plan information 20 via the input unit 320. The generation unit 354 also acquires search results from the search unit 353. The generation unit 354 generates three-dimensional space information based on the floor plan information 20 and the search results of the search unit 353. The generation unit 354 outputs the three-dimensional space information to the adjustment unit 355. Other details regarding the generation unit 354 are the same as those described with reference to FIG. 14 .

[0140] The adjustment unit 355 outputs the three-dimensional spatial information acquired from the generation unit 354 to the display unit 330 for display. Furthermore, when the adjustment unit 355 acquires adjustment parameters from the adjusted query extraction unit 352, the adjustment unit 355 adjusts the three-dimensional spatial information using the adjustment parameters, and outputs the adjusted three-dimensional spatial information to the display unit 330 for display. Other descriptions regarding the adjustment unit 355 are the same as those regarding the adjustment unit 355 described in FIG. 14 .

[0141] The prompt adjustment unit 400 receives the user information 10a and the text information 10b, generates an adjustment prompt, and outputs the generated adjustment prompt to the adjustment query extraction unit 352. Furthermore, if the prompt adjustment unit 400 receives the text information 10c after the three-dimensional space information is displayed on the display unit 330, it generates an adjustment prompt and outputs it to the adjustment query extraction unit 352. Other explanations regarding the prompt adjustment unit 400 are the same as those regarding the prompt adjustment unit 400 described using FIG. 15 .

[0142] Next, an example of a processing procedure of the information processing device 300 according to the present embodiment will be described. Fig. 18 and Fig. 19 are flowcharts showing the processing procedure of the information processing device according to the present embodiment. As shown in Fig. 18, the information processing device 300 receives input of user information 10a, text information 10b, and floor plan information 20 (step S301).

[0143] The attribute extraction unit 351 of the information processing device 300 extracts attribute information based on the user information 10a (step S302). The prompt adjustment unit 400 of the information processing device 300 generates an adjustment prompt based on the user information 10a and the text information 10b (step S303).

[0144] The adjustment query extraction unit 352 of the information processing device 300 extracts preference information based on the adjustment prompt (step S304). The search unit 353 of the information processing device 300 identifies a vector corresponding to the attribute based on the attribute information and the vector DB 341 (step S305). The search unit 353 searches for asset information from the asset DB 342 based on the vector corresponding to the attribute (step S306).

[0145] The search unit 353 updates the parameters of the assets included in the asset information based on the preference information (step S307). The generation unit 354 of the information processing device 300 generates three-dimensional space information based on the search result of the search unit 353 and the floor plan information 20 (step S308), and proceeds to step S309 in FIG. 19 .

[0146] The description will now move on to Fig. 19. The adjustment unit 355 of the information processing device 300 causes the display unit 330 to display the three-dimensional space information (step S309).

[0147] When new text information is received (Yes at step S310), the prompt adjustment unit 400 generates an adjustment prompt based on the already acquired user information 10a and the newly received text information 10c (step S311).

[0148] The adjustment query extraction unit 352 extracts adjustment parameters based on the adjustment prompt (step S312). The adjustment query extraction unit 352 generates adjustment parameters based on the preference information (step S313). The adjustment unit 355 updates the three-dimensional space information based on the adjustment parameters (step S314), and the process proceeds to step S309.

[0149] On the other hand, if new text information has not been received (step S310, No), information processing device 300 ends the process.

[0150] Next, the effects of the information processing device 300 according to the third embodiment will be described. The information processing device 300 receives input from a user (user information 10a, text information 10b, etc.), searches for assets in a three-dimensional space using the user's attributes and a predetermined language model, and generates three-dimensional space information. This makes it possible to easily generate a complete three-dimensional space even if the user inputs vague instructions.

[0151] For example, the information processing device 300 adds business knowledge information to the text information 10b based on a role identified from the user information 10a and aggregating multiple attributes. The information processing device 300 uses a predetermined language model (e.g., a private LLM) to extract preference information related to the user's preferences and attribute information related to the user's multiple attributes from the input information to which the business knowledge information has been added and the user information. The information processing device 300 generates an object (asset information) in a three-dimensional space based on the attribute information. The information processing device 300 updates the object based on the preference information and generates three-dimensional space information based on the updated object. This makes it possible to easily generate a complete three-dimensional space by supplementing the text information 10b using a role estimated from the user information, even if the user inputs an ambiguous instruction.

[0152] The information processing device 300 identifies a role that aggregates multiple attributes of the user based on the user information, and searches for business knowledge information corresponding to the role from the individual company knowledge DB 241, which associates roles with business knowledge information. This makes it possible to search for information that is insufficient when generating three-dimensional spatial information using text information alone.

[0153] When new text information 10c is received after the information processing device 300 displays the three-dimensional space information on the display unit 330, the information processing device 300 generates an adjustment prompt by adding business knowledge information to the text information 10c, and recreates the three-dimensional space information based on the adjustment prompt. This allows the three-dimensional space information to be closer to the information preferred by the user.

[0154] The above-described processing of the information processing device 300 is an example, and the information processing device 300 may execute other processing. For example, when the content included in the text information 10b and the content of the business knowledge information identified from the user information 10a are different for the parameters of the same asset, the information processing device 300 may provide feedback to the user and request that the user re-enter the text information 10b.

[0155] For example, if the text information 10b is "I want the wallpaper in my room to be blue," and the content of the business knowledge information identified from the user information 10a is "White wallpaper is preferable in my room," the prompt adjustment unit 400 determines that the color parameters for the same asset (wallpaper) are different. The difference in parameters indicates that the user's desires and the user's superior's desires regarding the room (three-dimensional space information) environment differ. In this case, the prompt adjustment unit 400 determines that the color parameters for the same asset are different (e.g., satisfy a predetermined condition) and requests the display unit 330 to re-enter the text information. For example, the prompt adjustment unit 400 may display a message such as "Since the business knowledge information is "White wallpaper is preferable in my room," please re-enter the text information" on the display unit 330 to prompt the user to re-enter the text information. This allows for the generation of a complete three-dimensional space optimal for the user and the user's role.

[0156] Furthermore, the information processing device 300 according to the third embodiment can be used in a conference where multiple users participate. For example, when a user inputs text information or audio information into the information processing device 300, the entire three-dimensional space envisioned by the user can be visually displayed to other users. This allows for efficient pre-storming and other activities.

[0157] Qualitative interviews are conducted to gather information for marketing, and there are conventional techniques for automating interviews. For example, one such conventional technique generates answers based on reference information using Retrieval-Augmented Generation (RAG) or similar to provide basic knowledge about the interview. However, with such conventional techniques, the answer information is influenced by the reference information, making it impossible to obtain information other than the reference information. Furthermore, if a generation AI is simply made to read past interview minutes and respond based on those minutes, it will end up repeating the answers.

[0158] To address the issues with the conventional technology described above, there is a need to apply an LLM-based Agent to obtain information (in-depth answers) other than the attached information that can be used for marketing, while utilizing additional information such as personas and product-related information.

[0159] An example of processing by an information processing device according to the fourth embodiment will be described. The information processing device according to the fourth embodiment will be referred to as an "information processing device 800." The information processing device 800 sets an LLM-based interviewer agent and has the interviewer agent apply interview tips and knowledge to generate in-depth questions that avoid devolving into irrelevant discussions. The information processing device 800 also sets an LLM-based interviewee agent and has the interviewee agent generate answers based on constraints that assume in-depth discussion, thereby obtaining in-depth answers related to latent needs and customer insights. The LLM used in the fourth embodiment is Chat-4, etc.

[0160] 20 is a diagram for explaining the processing of the information processing device according to the fourth embodiment. As shown in FIG.

[0161] The in-depth response generation unit 851 sets an LLM-based interviewer Agent 851a as the "interviewer." The interviewer Agent 851a generates question information for the interviewee based on the persona-related information 51a of the interviewer, the trend information 70a, the product-related information 70b, and the survey content information 70c stored in the storage unit 840.

[0162] The in-depth response generation unit 851 sets an LLM-based interviewee Agent 851b as the "interviewee." The interviewee Agent 851b generates answer information to the question information received from the interviewer based on the persona-related information 51b of the interviewee, the trend information 70a, the product-related information 70b, and the survey content information 70c stored in the storage unit 840.

[0163] For example, the trend information 70a includes information on popular products by gender and age group. The product-related information 70b includes product prices, product rankings, etc. The survey content information 70c includes an interview history. The interview history associates question information with answer information for the question information, which is information on answers by gender and age group.

[0164] For example, when a person 54 inputs input information 54a to the in-depth response generation unit 851 of the information processing device 800, the in-depth response generation unit 851 uses the input information 54a as a starting point (first question) to have an exchange between the interviewer and the interviewee, and outputs the interview result 56. The input information includes product information, main questions, etc.

[0165] Alternatively, instead of the interviewer Agent 851a, the person 55a may generate question information and ask the interviewee a question. Similarly, instead of the interviewee Agent 851b, the person 55b may refer to the question information and generate answer information.

[0166] For example, combinations of interviewers and interviewees include a pair of interviewer Agent 851a and person 55b, a pair of person 55a and interviewee Agent 851b, and a pair of interviewer Agent 851a and interviewee Agent 851b.

[0167] FIG. 21 is a diagram (1) showing an example of an interaction between an interviewer and an interviewee. FIG. 21 shows an interaction between person 55a and interviewee Agent 851b. For example, when person 55a (or person 54) asks an "initial question," interviewee Agent 851b makes an "initial response." When person 55a asks a "next question," interviewee Agent 851b makes a "probing response."

[0168] In the example described with reference to FIG. 21, information related to potential needs and customer insights can be obtained from the interviewee Agent 851b through in-depth investigation.

[0169] FIG. 22 is a diagram (2) showing an example of an interaction between an interviewer and an interviewee. FIG. 22 shows an interaction between an interviewer Agent 851a and an interviewee Agent 851b. For example, when person 54 asks an "initial question," interviewee Agent 851b makes an "initial response." When interviewer Agent 851a asks a "next question," interviewee Agent 851b makes an "in-depth response."

[0170] In the example illustrated in Figure 22, the interviewer and interviewee are automated, allowing for in-depth responses to be obtained automatically, and long exchanges can be continued in accordance with the personas without digressing into unrelated discussions.

[0171] Next, an example of prompt information set in the interviewer agent 851a will be described. Fig. 23 is a diagram showing an example of prompt information set in the interviewer agent. For example, the prompt information 52a set in the interviewer agent 851a includes prompt information 52a-1, 52a-2, and 52a-3.

[0172] The prompt information 52a-1 contains an outline of Chain of Thought (CoT) prompting. For example, among the information included in the prompt information 52a-1, "Please identify elements from the interviewee's answers that contribute to deeper exploration" and "Please set / update the direction of the questions" are information for setting axes to prevent questions from diverging in various directions. "Based on the direction of the questions, propose questions that contribute to the fundamental needs that lead to the interviewee's purchasing behavior. The trick is to alternate between asking about actions and thoughts" is information that shows tips for interviews aimed at deeper exploration, instructing the interviewer to ask about thoughts, then ask about related actions, and then ask about the motivation for those actions and the feelings at the time of those actions.

[0173] The prompt information 52a-2 sets rules for creating questions, such as "adjust questions to be as relevant to the interviewee as possible" to prevent the discussion from continuing to develop into unrelated topics.

[0174] The prompt information 52a-3 includes a plurality of examples of probing questions, the examples of which will not be shown here.

[0175] Next, an example of prompt information set for the interviewee agent 851b will be described. Fig. 24 is a diagram showing an example of prompt information set for the interviewee agent. For example, the prompt information 52b set for the interviewee agent 851b includes prompt information 52b-1, 52b-2, and 52b-3.

[0176] The prompt information 52b-1 sets an outline of the CoT prompting. Among the information included in the prompt information 52b-1, "Please imitate the person given as persona information" and "Please interpret the context, social situation, and general consumption trends from the perspective of the persona" are information that instructs the user to answer from the perspective of the given persona. "Please generate an answer in accordance with the rules, taking into account the conversation history" is information that refers to the conversation history and restricts the answer with rules to prevent duplicate answers.

[0177] The prompt information 52b-2 contains rules for creating answers to questions. For example, "Do not repeat what you have said before" and "When referring to similar information, please mention the motivation and underlying desire for the information" are rules for generating in-depth answers by allowing reference to similar content but requiring the addition of new information.

[0178] Reference information is set in the prompt information 52b-3. The reference information is information intended to provide information on which the interviewee's answers are based or as a base. For example, the reference information may include persona information, information on the interviewee and information related to the interviewee, information on social conditions, general consumption trends, questions from the interviewer, chat history, etc.

[0179] For example, the persona information is information shown in Fig. 25. Fig. 25 is a diagram showing an example of persona information. As shown in Fig. 25, the persona information 53b includes information such as name, gender, age, occupation, address, age, hobbies, lifestyle, family structure, values, consumption behavior, reliable information sources, and desired future.

[0180] An example of the results of a depth interview with interviewee Agent 851b for which persona information 53b shown in Figure 25 has been set is shown in Figure 26. Figure 26 is a diagram showing an example of the results of a depth interview. As shown in Figure 26, when person 54 asks an initial question 54-1, interviewee Agent 851b generates answer information 54-2 in response to initial question 54-1.

[0181] The interviewer Agent 851a generates question information 54-3 in response to the answer information 54-2. The interviewee Agent 851b generates answer information 54-4 in response to the question information 54-3. The interviewer Agent 851a generates question information 54-5 in response to the answer information 54-4. The interviewee Agent 851b generates answer information 54-6 in response to the question information 54-5.

[0182] In the response information shown in FIG. 26, the sentences in bold correspond to information related to latent needs and customer insights (deep-dive responses).

[0183] Next, a description will be given of an example configuration of the above-mentioned information processing device 800. Fig. 27 is a functional block diagram showing the configuration of the information processing device according to Example 4. As shown in Fig. 27, the information processing device 800 includes a communication unit 810, an input unit 820, a display unit 830, a storage unit 840, and a control unit 850.

[0184] The communication unit 810 executes data communication with external devices via a network, and is realized by a NIC or the like.

[0185] The input unit 820 is an input device that inputs various types of information to the information processing device 800. For example, the person 54 may operate the input unit 820 to input an initial question as described with reference to FIG.

[0186] The display unit 830 is a display device that displays information output from the control unit 850. For example, the display unit 830 displays the interview history as described with reference to FIG.

[0187] The storage unit 840 has various types of information used for executing processing by the in-depth response generation unit 851. For example, the storage unit 840 has trend information 70a, product-related information 70b, survey content information 70c, persona-related information 51a and 51b described in Fig. 20, prompt information described in Figs. 23 and 24, persona information 53b described in Fig. 25, etc. The storage unit 840 is a memory or the like.

[0188] The control unit 850 includes an in-depth response generation unit 851. The control unit 850 is realized by, for example, a CPU or an MPU. The control unit 850 may also be implemented by, for example, an integrated circuit such as an ASIC or an FPGA.

[0189] The in-depth response generation unit 851 sets an LLM-based interviewer agent and has the interviewer agent apply interview tips and knowledge to generate in-depth questions without devolving into unrelated discussions. The in-depth response generation unit 851 also sets an LLM-based interviewee agent and has the interviewee agent generate answers based on constraints that assume in-depth discussion, thereby obtaining in-depth answers related to latent needs and customer insights. The in-depth response generation unit 851 outputs and displays the interview results 56 on the display unit 830. Other descriptions of the in-depth response generation unit 851 correspond to the processes described in FIGS. 20 to 26.

[0190] Next, an example of a processing procedure of the information processing device 800 according to the fourth embodiment will be described. Fig. 28 is a flowchart showing the processing procedure of the information processing device according to the fourth embodiment. The in-depth response generation unit 851 of the information processing device 100 sets an interviewer Agent 851a (step S401). The in-depth response generation unit 851 sets an interviewee Agent 851b (step S402).

[0191] The in-depth response generation unit 851 receives the first question from the person 54 (step S403). The in-depth response generation unit 851 repeatedly causes the interviewer Agent 851a to generate question information and the interviewee Agent 851b to generate answer information (step S404). The in-depth response generation unit 851 causes the display unit 830 to output the interview results (step S405).

[0192] As described above, the information processing device 800 sets up an LLM-based interviewer agent and has the interviewer agent apply interview tips and knowledge to generate probing questions without devolving into irrelevant discussions. The information processing device 800 also sets up an LLM-based interviewee agent and has the interviewee agent generate answers based on constraints that assume probing, thereby obtaining probing answers related to latent needs and customer insights. This makes it possible to obtain probing answers that are useful for marketing purposes, etc., while conforming to attached information such as personas. Furthermore, it is possible to fully automate interviews that previously had to be conducted manually.

[0193] Traditionally, when qualitative interviews are conducted manually to gather information for marketing, it takes time and money to gather interviewees. Furthermore, when using a generative AI to play the role of an interviewee, it is possible to have the AI ​​read past interview minutes and respond, but this results in repeating the answers.

[0194] To address the above issues, it is necessary to give generative AI multiple personalities and roles, and through discussions between them, to review products and generate ideas from multiple perspectives.

[0195] The interview format includes a “depth interview” and a “group interview.” Figure 29 is a diagram for explaining the characteristics of a depth interview and a group interview.

[0196] In-depth interviews are conducted with one person as the subject, with the objective of digging deeper into the customer's decision-making process and gaining a detailed understanding. Suitable products are those that are directly related to daily life or that are difficult to discuss in public. The advantage is that it allows for in-depth and thorough information to be collected based on individual lifestyles. The disadvantage is that there is a possibility of bias depending on the content of the interview.

[0197] In group interviews, the number of subjects is "multiple people," the purpose is "to verify differences in reactions between groups and understand the reasons by attribute," and suitable products are "products used by multiple people, products that spark lively conversation, etc." The advantage is that "conversation can lead to a wealth of opinions and divergence of opinions." The disadvantage is "there is a possibility of bias depending on the participants."

[0198] An example of processing by an information processing device according to this embodiment 5 will be described. The information processing device according to this embodiment 5 will be referred to as an "information processing device 900." The information processing device 900 gives the generation AI multiple personalities and roles, and allows the generation AI to review products and generate ideas from multiple perspectives through discussions between them.

[0199] When constructing a personality, the information processing device 900 prepares items expected of a persona and automatically generates prompts indicating a user persona for the items. The items expected of a persona include age, family structure, personality, occupation, hobbies, lifestyle, values, consumption behavior, reliable information sources, and desired future.

[0200] When constructing a personality, the information processing device 900 has the function of extracting items expected of an interview persona from the contents of past interpersonal interviews, and forms a personality using generation AI based on the extraction results.

[0201] When configuring multiple virtual personalities, the information processing device 900 selects multiple personas generated by the generation AI. For example, when selecting multiple personas, the information processing device 900 selects personas so that there is no overlap in age, etc.

[0202] 30 is a diagram (1) for explaining the processing of the information processing device according to the fifth embodiment. In FIG. 30, a case where a group interview is performed will be explained.

[0203] For example, the information processing device 900 has a persona item table 941 , a persona prompt generation unit 951 , and an interview execution unit 952 .

[0204] The persona item table 941 is a table for storing information on items expected of a persona, such as age, family structure, personality, occupation, hobbies, lifestyle, values, consumption behavior, reliable information sources, and desired future.

[0205] The persona prompt generation unit 951 generates an interviewer prompt and prompts for multiple virtual customers (interviewees) based on the persona item table 941. In the example shown in Fig. 30, the persona prompt generation unit 951 generates an interviewer role definition prompt 81a as the interviewer prompt.

[0206] For example, the persona prompt generator 951 may utilize the LLM to generate prompts for the interviewer and the virtual customer.

[0207] The persona prompt generation unit 951 generates virtual customer persona definition prompts 81b, 81c, and 81d as prompts for virtual customers.

[0208] The interview execution unit 952 sets an LLM-based interviewer Agent 952a as the "interviewer." The interviewer Agent 952a generates question information for the virtual customer based on the interviewer Role definition Prompt 81a. For example, the interviewer Agent 952a generates the initial question information using an externally specified virtual product description and the initial discussion content.

[0209] The interview execution unit 952 sets LLM-based virtual customer Agents 952b, 952c, and 952d as "virtual customers (interviewees)." The virtual customer Agent 952b generates answer information to the question information based on the virtual customer persona definition Prompt 81b. The virtual customer Agent 952b generates answer information to the question information based on the virtual customer persona definition Prompt 81c. The virtual customer Agent 952c generates answer information to the question information based on the virtual customer persona definition Prompt 81c. The virtual customer Agent 952d generates answer information to the question information based on the virtual customer persona definition Prompt 81d.

[0210] The interview execution unit 952 executes a group interview based on an interviewer Agent 952a and virtual customer Agents 952b, 952c, and 952d, and outputs the interview results.

[0211] 31 is a diagram showing the results of a group interview. For example, the persona of virtual customer Agent 952b is "Name: Masao, Gender: Male, Age: 60s, Personality: Humorous." The persona of virtual customer Agent 952c is "Name: Ryo, Gender: Male, Age: 20s, Personality: Adventurous." The persona of virtual customer Agent 952d is "Name: Maki, Gender: Female, Age: 30s, Personality: Sensitive."

[0212] In the group interview in Figure 31, the virtual product description and situation specified externally was, "We have developed a new canned beer. The name of the canned beer is 'Autumn Blessing Ale' and it has an alcohol content of 5%. The can has a picture of autumn leaves on it. The interviewer distributed a beer to each of the three users and they were allowed to drink it however they liked."

[0213] When the interviewer Agent 952a presents question information 81-1 (initial question information), the virtual customer Agent 952b generates answer information 81-2, the virtual customer Agent 952c generates answer information 81-3, and the virtual customer Agent 952d generates answer information 81-4. The interviewer Agent 952a generates question information 81-5 in response to the answer information 81-2, 81-3, and 81-4.

[0214] The results of the group interview shown in Figure 31 generated information from a variety of perspectives, such as product design and usage scenarios (food that goes well with the product).

[0215] A case where a group interview is performed has been described with reference to Fig. 30. Fig. 32 is a diagram (2) for explaining the processing of the information processing device according to the fifth embodiment. Fig. 32 describes a case where a depth interview is performed.

[0216] The persona item table 941, persona prompt generation unit 951, and interview execution unit 952 are the same as those described in FIG.

[0217] For example, the persona prompt generation unit 951 generates an interviewer role definition prompt 81a as a prompt for the interviewer, and a virtual customer role definition prompt 81e as a prompt for the virtual customer.

[0218] The interview execution unit 952 sets an LLM-based interviewer Agent 952a as the "interviewer." The interviewer Agent 952a generates question information for the virtual customer based on the interviewer Role definition Prompt 81a. For example, the interviewer Agent 952a generates the initial question information using an externally specified interview question.

[0219] The interview execution unit 952 sets an LLM-based virtual customer Agent 952e as a "virtual customer (interviewee)." The virtual customer Agent 952e generates answer information to the question information based on the virtual customer Role definition Prompt 81e.

[0220] In the example of processing shown in Fig. 32, product knowledge 942 is provided to the virtual customer Agent 952e. Fig. 33 is a diagram showing an example of product knowledge. As shown in Fig. 33, the product knowledge 942 includes items such as product name, alcohol content, price, target demographic, TV commercial summary, and product design, and content corresponding to each item is set. The virtual customer Agent 952e uses the product knowledge 942 to generate answer information to the question information.

[0221] The interview execution unit 952 executes an in-depth interview based on an interviewer Agent 952a and a virtual customer Agent 952e, and outputs the interview results.

[0222] Figure 34 shows the results of the in-depth interview. For comparison, Figure 34 also shows the response information generated by "general LLM only." The first externally specified question is "What image did you have of 'Autumn Flavor' before purchasing?"

[0223] The answer information for only the general-purpose LLM is answer information 82-1. According to answer information 82-1, product knowledge is not provided, and therefore the answer is not on point.

[0224] The answer information of the virtual customer Agent 952e using the product knowledge 942 is answer information 82-2. According to the answer information 82-2, although some points of view are not a complete copy of the information provided, most of the answer information is output as is.

[0225] The answer of the interviewer Agent 952a (answer to the answer information 82-2) is answer information 82-3. According to the answer information 82-3, information on the "perspective" (not a verbatim copy) is extracted from the answer of the virtual customer.

[0226] According to the interview results shown in Figure 34, answer information based on product knowledge was obtained, and what is particularly important is that by using multiple agents, virtual customers and interviewers, it is possible to present new insights that break down the given product knowledge 942.

[0227] Next, a description will be given of an example configuration of the above-mentioned information processing device 900. Fig. 35 is a functional block diagram showing the configuration of an information processing device according to Example 5. As shown in Fig. 35, the information processing device 900 includes a communication unit 910, an input unit 920, a display unit 930, a storage unit 940, and a control unit 950.

[0228] The communication unit 910 executes data communication with external devices via a network, and is realized by a NIC or the like.

[0229] The input unit 920 is an input device that inputs various types of information to the information processing device 900. For example, the person 54 may operate the input unit 920 to input an interview question such as that described with reference to FIG.

[0230] The display unit 930 is a display device that displays information output from the control unit 950. For example, the display unit 930 displays the interview history as described with reference to FIG.

[0231] The storage unit 940 has various information used by the persona prompt generation unit 951, the interview execution unit 952, etc. to execute processing. For example, the storage unit 940 has a persona item table 941, product knowledge 942, etc. The storage unit 940 is a memory or the like.

[0232] The control unit 950 includes a persona prompt generation unit 951 and an interview execution unit 952. The control unit 950 may be implemented by an integrated circuit such as an ASIC or FPGA.

[0233] The persona prompt generation unit 951 generates prompts for an interviewer and prompts (including persona information) for multiple virtual customers (interviewees) based on the persona item table 941. The persona prompt generation unit 951 may generate prompts for the interviewer and the virtual customers using the LLM.

[0234] When selecting multiple personas, the persona prompt generation unit 951 selects personas so that there is no overlap in age groups, etc., and sets the selected personas as prompts for the virtual customers.

[0235] When conducting a group interview, the interview execution unit 952 sets an LLM-based interviewer Agent as the "interviewer" and sets multiple LLM-based virtual customer Agents as "virtual customers (interviewees)." The interview execution unit 952 conducts the group interview based on the set information and outputs the interview results. The process by which the interview execution unit 952 conducts a group interview corresponds to the process described in FIG. 30.

[0236] When the interview execution unit 952 executes a depth interview, it sets an LLM-based interviewer Agent as the "interviewer" and an LLM-based virtual customer Agent as the "virtual customer (interviewee)." The interview execution unit 952 executes a depth interview based on the set information and outputs the interview results. The process by which the interview execution unit 952 executes a depth interview corresponds to the process described in FIG. 32 .

[0237] Next, an example of a processing procedure of the information processing device 900 according to this embodiment 5 will be described. Fig. 36 is a flowchart showing the processing procedure of the information processing device according to this embodiment 5. As shown in Fig. 32, the persona prompt generation unit 951 of the information processing device 900 generates an interviewer prompt and a virtual customer prompt based on the persona item table 941 (step S501).

[0238] The interview execution unit 952 of the information processing device 900 sets an interviewer Agent (step S502), and then sets an interviewee Agent (step S503).

[0239] The interview execution unit 952 receives the virtual product description and the initial discussion content (step S504). The interview execution unit 952 repeatedly executes a process of having the interviewer Agent generate question information and the interviewee Agent generate answer information (step S505). The interview execution unit 952 outputs the interview results to the display unit 930 (step S506).

[0240] As described above, the information processing device 900 according to the fifth embodiment allows the generation AI to have multiple personalities and roles, and by giving the generation AI multiple personalities and roles, product reviews and idea generation can be performed from multiple perspectives through discussions between them.

[0241] Next, an example of a hardware configuration of a computer that realizes the same functions as the information processing devices 100, 200, 300, 800, and 900 shown in the above embodiments will be described in order.

[0242] FIG. 37 is a diagram illustrating an example of the hardware configuration of a computer that realizes the same functions as the information processing apparatus of the first embodiment. As illustrated in FIG. 37, the computer 500 includes a CPU 501 that executes various arithmetic processes, an input device 502 that accepts data input from a user, and a display 503. The computer 500 also includes a communication device 504 that exchanges data with external devices via a wired or wireless network, and an interface device 505. A microphone, a speaker, and the like may be connected to the interface device 505. The computer 500 also includes a RAM 506 that temporarily stores various information, and a hard disk drive 507. The devices 501 to 507 are connected to a bus 508.

[0243] The hard disk drive 507 stores an attribute extraction program 507 a, an adjusted query extraction program 507 b, a search program 507 c, a generation program 507 d, and an adjustment program 507 e. The CPU 501 reads out each of the programs 507 a to 507 e and loads them into the RAM 506.

[0244] The attribute extraction program 507a functions as the attribute extraction process 506a. The adjusted query extraction program 507b functions as the adjusted query extraction process 506b. The search program 507c functions as the search process 506c. The generation program 507d functions as the generation process 506d. The adjustment program 507e functions as the adjustment process 506e.

[0245] The processing of the attribute extraction process 506a corresponds to the processing of the attribute extraction unit 151. The processing of the adjusted query extraction process 506b corresponds to the processing of the adjusted query extraction unit 152. The processing of the search process 506c corresponds to the processing of the search unit 153. The processing of the generation process 506d corresponds to the processing of the generation unit 154. The processing of the adjustment process 506e corresponds to the processing of the adjustment unit 155.

[0246] It should be noted that each of the programs 507a to 507e does not necessarily have to be stored in the hard disk drive 507 from the beginning. For example, each program may be stored in a "portable physical medium" such as a flexible disk (FD), CD-ROM, DVD, magneto-optical disk, or IC card that is inserted into the computer 500. Then, the computer 500 may read and execute each of the programs 507a to 507e.

[0247] FIG. 38 is a diagram showing an example of the hardware configuration of a computer that realizes the same functions as the information processing apparatus of the second embodiment. As shown in FIG. 38, the computer 600 includes a CPU 601 that executes various arithmetic processes, an input device 602 that accepts data input from a user, and a display 603. The computer 600 also includes a communication device 604 that exchanges data with external devices via a wired or wireless network, and an interface device 605. A microphone, a speaker, and the like may be connected to the interface device 605. The computer 600 also includes a RAM 606 that temporarily stores various information, and a hard disk drive 607. The devices 601 to 607 are connected to a bus 608.

[0248] The hard disk drive 607 stores a role identification program 607 a, a pre-processing program 607 b, a search program 607 c, a post-processing program 607 d, a response generation program 607 e, and an image generation program 607 f. The CPU 601 reads out each of the programs 607 a to 607 f and loads them into the RAM 606.

[0249] The role identification program 607a functions as a role identification process 606a. The pre-processing program 607b functions as a pre-processing process 606b. The search program 607c functions as a search process 606c. The post-processing program 607d functions as a post-processing process 606d. The answer generation program 607e functions as an answer generation process 606e. The image generation program 607f functions as an image generation process 606f.

[0250] The processing of the role identification process 606a corresponds to the processing of the role identification unit 251. The processing of the pre-processing process 606b corresponds to the processing of the pre-processing unit 252. The processing of the search process 606c corresponds to the processing of the search unit 253. The processing of the post-processing process 606d corresponds to the processing of the post-processing unit 254. The processing of the answer generation process 606e corresponds to the processing of the answer generation unit 255. The processing of the image generation process 606f corresponds to the processing of the image generation unit 256.

[0251] It should be noted that each of the programs 607a to 607f does not necessarily have to be stored in the hard disk drive 607 from the beginning. For example, each program may be stored in a "portable physical medium" such as a flexible disk (FD), CD-ROM, DVD, magneto-optical disk, or IC card that is inserted into the computer 600. Then, the computer 600 may read and execute each of the programs 607a to 607f.

[0252] FIG. 39 is a diagram showing an example of the hardware configuration of a computer that realizes the same functions as the information processing apparatus of the third embodiment. As shown in FIG. 39, the computer 700 includes a CPU 701 that executes various arithmetic processes, an input device 702 that accepts data input from a user, and a display 703. The computer 700 also includes a communication device 704 that exchanges data with external devices via a wired or wireless network, and an interface device 705. A microphone, a speaker, and the like may be connected to the interface device 705. The computer 700 also includes a RAM 706 that temporarily stores various information, and a hard disk drive 707. The devices 701 to 707 are connected to a bus 708.

[0253] The hard disk drive 707 stores an attribute extraction program 707 a, an adjusted query extraction program 707 b, a search program 707 c, a generation program 707 d, an adjustment program 707 e, and a prompt adjustment program 707 f. The CPU 701 reads out each of the programs 707 a to 707 f and loads them into the RAM 706.

[0254] The attribute extraction program 707a functions as the attribute extraction process 706a. The adjusted query extraction program 707b functions as the adjusted query extraction process 706b. The search program 707c functions as the search process 706c. The generation program 707d functions as the generation process 706d. The adjustment program 707e functions as the adjustment process 706e. The prompt adjustment program 707f functions as the prompt adjustment process 706f.

[0255] The processing of the attribute extraction process 706a corresponds to the processing of the attribute extraction unit 351. The processing of the adjusted query extraction process 706b corresponds to the processing of the adjusted query extraction unit 352. The processing of the search process 706c corresponds to the processing of the search unit 353. The processing of the generation process 706d corresponds to the processing of the generation unit 354. The processing of the adjustment process 706e corresponds to the processing of the adjustment unit 355. The processing of the prompt adjustment process 706f corresponds to the processing of the prompt adjustment unit 400.

[0256] It should be noted that each of the programs 707a to 707f does not necessarily have to be stored in the hard disk drive 707 from the beginning. For example, each program may be stored in a "portable physical medium" such as a flexible disk (FD), CD-ROM, DVD, magneto-optical disk, or IC card that is inserted into the computer 700. Then, the computer 700 may read and execute each of the programs 707a to 707f.

[0257] FIG. 40 is a diagram showing an example of the hardware configuration of a computer that realizes the same functions as the information processing apparatus of the fourth embodiment. As shown in FIG. 39, the computer 1000 includes a CPU 1001 that executes various arithmetic processes, an input device 1002 that accepts data input from a user, and a display 1003. The computer 1000 also includes a communication device 1004 that exchanges data with external devices via a wired or wireless network, and an interface device 1005. A microphone, a speaker, and the like may be connected to the interface device 1005. The computer 1000 also includes a RAM 1006 that temporarily stores various information, and a hard disk drive 1007. The devices 1001 to 1007 are connected to a bus 1008.

[0258] The hard disk drive 1007 stores a deep analysis response generation program 1007 a. The CPU 1001 reads out the deep analysis response generation program 1007 a and loads it into the RAM 1006 .

[0259] The in-depth response generation program 1007a functions as an in-depth response generation process 1006a. The processing of the in-depth response generation process 1006a corresponds to the processing of the in-depth response generation unit 851.

[0260] It should be noted that the in-depth response generation program 1007a does not necessarily have to be stored in the hard disk drive 1007 from the beginning. For example, each program may be stored in a "portable physical medium" such as a flexible disk (FD), CD-ROM, DVD, magneto-optical disk, or IC card that is inserted into the computer 1000. Then, the computer 1000 may read and execute the in-depth response generation program 1007a.

[0261] FIG. 41 is a diagram showing an example of the hardware configuration of a computer that realizes the same functions as the information processing apparatus of the fifth embodiment. As shown in FIG. 41, the computer 1100 includes a CPU 1101 that executes various arithmetic processes, an input device 1102 that accepts data input from a user, and a display 1103. The computer 1100 also includes a communication device 1104 that exchanges data with external devices via a wired or wireless network, and an interface device 1105. A microphone, a speaker, and the like may be connected to the interface device 1105. The computer 1100 also includes a RAM 1106 that temporarily stores various information, and a hard disk drive 1107. The devices 1101 to 1107 are connected to a bus 1108.

[0262] The hard disk drive 1107 stores a persona prompt generation program 1107a and an interview execution program 1107b. The CPU 701 reads out the programs 1107a and 1107b and loads them into the RAM 1106.

[0263] The persona prompt generation program 1107a functions as the persona prompt generation process 1106a, and the interview execution program 1107b functions as the interview execution process 1106b.

[0264] The processing of the persona prompt generation process 1106a corresponds to the processing of the persona prompt generation unit 951. The processing of the interview execution process 1106b corresponds to the processing of the interview execution unit 952.

[0265] It should be noted that the programs 1107a and 1107b do not necessarily have to be stored in the hard disk drive 1107 from the beginning. For example, the programs may be stored in a "portable physical medium" such as a flexible disk (FD), CD-ROM, DVD, magneto-optical disk, or IC card that is inserted into the computer 1100. The computer 1100 may then read and execute the programs 1107a and 1107b.

[0266] 100, 200, 300, 800, 900 Information processing device 110, 210, 310, 810, 910 Communication unit 120, 220, 320, 820, 920 Input unit 130, 230, 330, 830, 930 Display unit 140, 240, 340, 840, 940 Storage unit 141 Vector DB 142 Asset DB 150, 250, 350, 850, 950 Control unit 151, 351 Attribute extraction unit 152, 352 Adjustment query extraction unit 153, 353 Search unit 154, 354 Generation unit 155, 355 Adjustment unit 241 Individual company knowledge DB 251 Role identification unit 252 Preprocessing unit 253 Search unit 254 Post-processing unit 255 Answer generation unit 256 Image generation unit 400 Prompt adjustment unit 851 Deep dive response generation unit 951 Persona prompt generation unit 952 Interview execution unit

Claims

1. A generation program that causes a computer to execute the following processes: accept input from a user; generate objects in a three-dimensional space from the accepted input using the attributes of the user and a specified language model; and generate a three-dimensional space based on the objects in the three-dimensional space that have been generated.

2. The generation program of claim 1, characterized in that the receiving process receives as the input input information of the user and user information including multiple attributes of the user, and the generating process of the object in the three-dimensional space includes a process of adding business knowledge information to the input information based on a role identified from the user information and which aggregates the multiple attributes, a process of extracting preference information related to the user's preferences and attribute information related to the multiple attributes of the user from the input information to which the business knowledge information has been added and the user information using a predetermined language model, and a process of generating an object in the three-dimensional space based on the attribute information, and the generating process of the three-dimensional space includes a process of updating the object based on the preference information and generating the three-dimensional space based on the updated object.

3. The generation program according to claim 2, further comprising a process of identifying a role that aggregates multiple attributes of the user based on the user information, and searching for business knowledge information corresponding to the role identified by the identification process from a database that associates the roles with business knowledge information.

4. The generation program according to claim 2, further comprising the step of causing the computer to execute a process of, when new input information is received from the user, adding the business knowledge information to the new input information, updating the object based on the new input information to which the business knowledge information has been added, and recreating the three-dimensional space.

5. The generation program according to claim 2, further causing a computer to execute a process of providing feedback to said user when said user's input information satisfies a predetermined condition.

6. A generation program that causes a computer to execute the following processes: accept input information from a user and user information indicating the attributes of the user; use a predetermined language model to extract preference information related to the user's preferences and attribute information related to the user's attributes from the input information and the user information; generate an object in a three-dimensional space based on the attribute information; update the object based on the preference information; and generate a three-dimensional space based on the updated object.

7. A generation program that causes a computer to execute the following processes: accept user information including a user's input information and multiple attributes of the user; identify a role that aggregates the multiple attributes of the user based on the user information; search for business knowledge information corresponding to the role identified by the identification process from a database that associates the roles with business knowledge information; and generate a prompt to be input into a predetermined language model based on the input information and the business knowledge information searched for by the search process, the prompt being capable of outputting a response corresponding to the input information.

8. A method for generating a three-dimensional space, the method comprising the steps of: receiving an input from a user; generating an object in a three-dimensional space from the received input using the attributes of the user and a predetermined language model; and generating a three-dimensional space based on the object in the three-dimensional space that has been generated.

9. The method of generating an object in the three-dimensional space, characterized in that the receiving process receives as the input input information of the user and user information including multiple attributes of the user, and the process of generating an object in the three-dimensional space includes a process of adding business knowledge information to the input information based on a role identified from the user information and which aggregates the multiple attributes, a process of extracting preference information related to the user's preferences and attribute information related to the multiple attributes of the user from the input information to which the business knowledge information has been added and the user information using a predetermined language model, and a process of generating an object in the three-dimensional space based on the attribute information, and the process of generating the three-dimensional space includes a process of updating the object based on the preference information and generating the three-dimensional space based on the updated object.

10. The generation method described in claim 9, further characterized in that the computer executes a process of identifying a role that aggregates multiple attributes of the user based on the user information, and searching for business knowledge information corresponding to the role identified by the identification process from a database that associates the roles with business knowledge information.

11. The generation method according to claim 9, further comprising the step of: when new input information is received from the user, adding the business knowledge information to the new input information, updating the object based on the new input information to which the business knowledge information has been added, and recreating the three-dimensional space, wherein the computer further executes the process.

12. The generating method according to claim 9, further comprising the step of: providing feedback to said user when said input information from said user satisfies a predetermined condition, said computer further executing said processing.

13. A generation method characterized by having a computer execute the following processes: accepting input information from a user and user information indicating the attributes of the user; extracting preference information related to the user's preferences and attribute information related to the user's attributes from the input information and the user information using a predetermined language model; generating an object in a three-dimensional space based on the attribute information; updating the object based on the preference information; and generating a three-dimensional space based on the updated object.

14. A generation method characterized in that a computer executes the following processes: accepts user information including a user's input information and multiple attributes of the user; identifies a role that aggregates the multiple attributes of the user based on the user information; searches for business knowledge information corresponding to the role identified by the identification process from a database that associates the roles with business knowledge information; and generates a prompt to be input into a predetermined language model based on the input information and the business knowledge information searched for by the search process, the prompt being capable of outputting a response corresponding to the input information.

15. An information processing device having a control unit that executes the process of receiving input from a user, generating an object in a three-dimensional space from the received input using the attributes of the user and a predetermined language model, and generating a three-dimensional space based on the generated object in the three-dimensional space.

16. The information processing device described in claim 15, characterized in that the receiving process receives, as the input, input information of the user and user information including multiple attributes of the user, and the generating process of the object in the three-dimensional space includes a process of adding business knowledge information to the input information based on a role identified from the user information and which aggregates the multiple attributes, a process of extracting preference information related to the user's preferences and attribute information related to the multiple attributes of the user from the input information to which the business knowledge information has been added and the user information using a predetermined language model, and a process of generating an object in the three-dimensional space based on the attribute information, and the generating process of the three-dimensional space includes a process of updating the object based on the preference information and generating the three-dimensional space based on the updated object.

17. The information processing device according to claim 16, further characterized in that the control unit identifies a role that aggregates multiple attributes of the user based on the user information, and further executes a process of searching for business knowledge information corresponding to the role identified by the identification process from a database that associates the roles with business knowledge information.

18. The information processing device according to claim 16, characterized in that the control unit, when receiving new input information from the user, further executes a process of adding the business knowledge information to the new input information, updating the object based on the new input information to which the business knowledge information has been added, and recreating the three-dimensional space.

19. The information processing device according to claim 16, characterized in that the control unit further executes a process of providing feedback to the user when the information input by the user satisfies a predetermined condition.

20. An information processing device having a control unit that executes the following processes: accepts input information from a user and user information indicating the attributes of the user; extracts preference information regarding the user's preferences and attribute information regarding the user's attributes from the input information and the user information using a predetermined language model; generates an object in a three-dimensional space based on the attribute information; updates the object based on the preference information; and generates a three-dimensional space based on the updated object.

21. An information processing device having a control unit that executes the following processes: accepts user information including input information of a user and multiple attributes of the user; identifies a role that aggregates the multiple attributes of the user based on the user information; searches for business knowledge information corresponding to the role identified by the identification process from a database that associates the roles with business knowledge information; and generates a prompt to be input into a predetermined language model based on the input information and the business knowledge information searched for by the search process, the prompt being capable of outputting a response corresponding to the input information.

Images