Information processing apparatus, information processing method, and program

Abstract

To solve the problem that it is difficult to generate an appropriate prompt to obtain an answer desired by a user who is not familiar with a generative AI model.SOLUTION: A processor that determines selection policy information for determining a conversation partner based on the inputted information, and generates a prompt based on the selection policy information; and an output apparatus that is connected to a generative AI model that generates the conversation partner, and outputs the prompt.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present disclosure relates to an information processing device, an information processing method, and a program, and in particular to so-called Retrieval Augmented Generation (RAG), which combines retrieval and generative AI models (artificial intelligence). [Background technology]

[0002] When proposing a proposal or new idea, it is common to create a proposal, get approval or judgment from a superior, and obtain agreement from the relevant parties. In order to obtain the superior's judgment and agreement from the relevant parties, it is necessary to explain the usefulness of the contents of the proposal and the idea.

[0003] For this reason, before creating a proposal, it is common to have a brief conversation with a colleague about the merits and problems of the plan or idea, in order to improve it, known as brushing up. Talking with a colleague can also lead to reviewing the plan or idea from a new perspective. A conversation can also be called a discussion, consultation, debate, or argument.

[0004] Conversations with colleagues and superiors are not only for brushing up, but also for casual conversations and take place in a variety of situations. For example, when exchanging opinions with colleagues to efficiently complete one's assigned work, when a person in charge of interdepartmental collaboration receives information from a colleague who is knowledgeable about information in other departments, or when receiving ideas for project management from a senior colleague, etc. Conversations with colleagues and superiors in various situations often lead to the creation of new plans and ideas.

[0005] However, with the widespread adoption of teleworking, staggered work hours, and reduced working hours due to work style reforms, it can be difficult to have conversations with colleagues. On the other hand, many people believe that conversations are not worth setting up a meeting in advance, so when it is difficult to have conversations with colleagues, they tend to give up on the idea.

[0006] Recently, information processing devices have been proposed that utilize value-added application programming interfaces (APIs) based on large language models (LLMs). LLMs are generative artificial intelligence (AI) models specialized for language processing. When a user inputs a prompt, they generate a response and output it to the user.

[0007] Role-playing is being conducted with an AI customer role that has a set personality (for example, Non-Patent Document 1). [Prior art documents] [Non-patent literature]

[0008] [Non-Patent Document 1] Interactive Solutions Inc. iRolePlay<URL: https: / / www.interactive-solutions.co.jp / common / pdf / pr_20240411_01.pdf> Summary of the Invention [Problem to be solved by the invention]

[0009] When a user converses with a colleague, the user decides who to converse with and what the conversation will be about. In Non-Patent Document 1, since AI is trained to play the role of a "customer," the personality that is set is presumably the personality that the user wants to train or a randomly selected personality, and the content of the conversation is also determined by the AI, so the user cannot converse with the person they want and on the content they want. [Means for solving the problem]

[0010] An information processing device in one aspect of the present disclosure includes a processor that, when input information is input, determines selection policy information for determining a dialogue partner based on the input information and generates a prompt based on the selection policy information, and an output device connected to a generative AI model that generates the dialogue partner and outputs the prompt. [Effects of the Invention]

[0011] According to the present disclosure, when a dialogue is conducted using a generative AI model, an appropriate dialogue can be conducted. [Brief explanation of the drawings]

[0012] [Figure 1] Overall system configuration diagram [Figure 2] FIG. 1 shows a configuration of an information processing device. [Figure 3] Diagram showing the structure of the person database [Figure 4] FIG. 1 is a diagram showing a processing flow executed by an information processing device. [Figure 5] Diagram showing an example table [Figure 6] Diagram showing example prompts [Figure 7] Diagram showing an example of a dialogue DETAILED DESCRIPTION OF THE INVENTION

[0013] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings as appropriate.

[0014] (Embodiment 1) FIG. 1 shows the overall configuration of the system 100.

[0015] The system 100 includes an input / output terminal 110, an information processing device 120, a person database (DB) 130, and an LLM 140.

[0016] The input / output terminal 110 is a device into which request information is input from a user and into which conversations in response to the user's request information are input / output. The request information and conversations may be in the form of sentences (natural sentences) used in daily life.

[0017] The information processing device 120 is a device that generates a prompt to be output to the LLM 140 based on request information input by the user. When the request information input by the user from the input / output terminal 110 is input, the information processing device 120 uses the person DB 130 to select a colleague to converse with.

[0018] The person DB 130 is a storage device such as a hard disk drive (HDD) or a fixed memory (SSD: Solid State Drive).

[0019] The LLM 140 is a generative AI model specialized for language processing. The LLM 140 receives a prompt generated by the information processing device 120 and generates answer content in response to the input prompt. The LLM 140 can use any LLM, such as ChatGPT by OpenAI or tsuzumi by NTT (Nippon Telegraph and Telephone Corporation). The LLM 140 may be configured in the cloud or on a terminal. Because tsuzumi has a small parameter size, when tsuzumi is used, the LLM 140 can be configured on a terminal.

[0020] FIG. 2 shows the configuration of the information processing device 120.

[0021] The information processing device 120 includes a processor 121, an input device 122, and an output device 123. In addition to the devices shown in FIG. 2, the information processing device 120 may include a communication device, a memory, and a storage device. The devices such as the processor 121 and the input device 122 are connected by a bus 124 for communicating information. The bus 124 may be configured using a single bus, or may be configured using different buses between the devices.

[0022] The processor 121 is composed of a computer including a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), etc., and is executed by a program. When dialogue request information is input from the input / output terminal 110, the processor 121 searches the person DB 130 based on the user who input the dialogue request information. The processor 121 determines dialogue partner selection policy information based on the search response obtained from the person DB 130, and generates a prompt to be output to the LLM 140. The processor 121 outputs the generated prompt to the LLM 140 via the output device 123. When the prompt generated by the processor 121 is output to the LLM 140, the LLM 140 generates a dialogue partner, and the user can have a conversation with the LLM 140 (the dialogue partner generated by the LLM 140).

[0023] The input device 122 may be an input device (e.g., a keyboard, a mouse, a microphone, a switch, a button, a sensor, etc.) to which request information is input from the outside, or may be an interface or communication device to which request information is input from the input / output terminal 110. The input device 122 may be connected to the knowledge DB 130 and the LLM 140.

[0024] The output device 123 may be an output device (for example, a display, a speaker, an LED lamp, etc.) that outputs the conversation to the outside, or may be an interface or communication device that transmits the conversation to the input / output terminal 110. The output device 123 may be connected to the knowledge DB 130 and the LLM 140.

[0025] The input device 122 and the output device 123 may be integrated (for example, a touch panel), or the input device 122 and / or the output device 123 may be a communication device.

[0026] The memory is a computer-readable recording medium and may be configured by, for example, at least one of a ROM (Read Only Memory), an EPROM (Erasable Programmable ROM), an EEPROM (Electrically Erasable Programmable ROM), a RAM (Random Access Memory), etc. The memory may also be called a register, a cache, a main memory (primary storage device), etc. The memory can store executable programs (program codes), software modules, etc. for implementing a wireless communication method according to an embodiment of the present disclosure.

[0027] The storage is a computer-readable recording medium, and may be, for example, at least one of an optical disk such as a CD-ROM (Compact Disc ROM), a hard disk drive, a flexible disk, a magneto-optical disk (e.g., a compact disk, a digital versatile disk, a Blue-ray® disk), a smart card, a flash memory (e.g., a card, a stick, a key drive), a floppy disk, a magnetic strip, etc. The storage may also be referred to as an auxiliary storage device. The above-mentioned storage medium may be, for example, a database, a server, or other suitable medium including at least one of a memory and a storage.

[0028] A communication device is hardware (input / output device) for communicating between computers via at least one of a wired network and a wireless network, and is also called, for example, a network device, a network controller, a network card, or a communication module.

[0029] The information processing device 120 may also be configured to include hardware such as a microprocessor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a programmable logic device (PLD), or a field programmable gate array (FPGA), and some or all of the functional blocks may be realized by the hardware. For example, the processor 121 may be implemented using at least one of these pieces of hardware.

[0030] FIG. 3 shows the configuration of the person DB 130.

[0031] The person DB 130 is a knowledge DB that stores information and tables related to the AI ​​model of a person (hereinafter simply referred to as "person") who will be a conversation partner. The person DB 130 may be provided on a terminal, or on a network or cloud. The person DB 130 may store information related to multiple people. Each of the multiple person DBs 130 may store information related to each person.

[0032] For example, person DB 130 may be configured with data for three people, namely, data 131 storing information about person A (hereinafter referred to as "person A data"), data 132 storing information about person B (hereinafter referred to as "person B data"), and data 133 storing information about person C (hereinafter referred to as "person C data"), and table 134. The number of people stored in person DB 130 does not have to be three. Table 134 stores information about people who are compatible with the user. Information about people who are incompatible with the user may also be stored in table 134.

[0033] The information about a person includes at least one item such as name, attributes, characteristics, personality, occupation, specialty, profession, values ​​information, emotions, and preference information.

[0034] The information about a person may be, for example, information based on a real person, information about a fictional person, or information about a historical figure. The attribute may be a relationship with the user, such as senior or junior, or may be gender or age. The characteristic may be, for example, execution ability, analytical ability, strong leadership, etc., or a role in a conversation. The role in a conversation may be a person who helps notice logical omissions or gaps, a person who makes suggestions that help people notice mistakes, a person who proposes ideas from a different perspective than other members, a person who asks questions about ideas proposed by other members to expand on them, a person who gives objective opinions on ideas presented in a conversation, etc. The personality may be, for example, quiet or talkative. The occupation may be, for example, engineer, marketing, etc. The value information may be information calculated by a value engine for each piece of information indicating a value, and the degree of the value may be indicated as a score. The values ​​may be, for example, compliance is important, sales are important, cost reduction is important, business flow is important, increasing the continuation rate of PDCA (Plan, Do, Check, Action) is important, and performance is important. The emotions may be values ​​output by an emotion engine, and may be values ​​obtained by analyzing either voice or facial expressions and outputting information indicating the degree of anger, sadness, surprise, or happiness as a score. The emotions may be output as a score obtained by rounding the output result of each emotion and displaying the ratio of positivity to negativity. The occupation may include characteristics of the occupation, such as a specialty architectural style in the case of an architect.

[0035] A fictional character may be prepared with specific information set regarding the set occupation, specialty, occupational characteristics, gender, attributes, etc., and a character who is compatible with the user may be determined from among multiple characters with specific information set.

[0036] (operation) The following describes the operation when a user interacts with a person generated by LLM140 before proposing an idea that the user has come up with to their boss for approval. FIG. 4 shows a processing flow executed by the information processing device 120.

[0037] The information processing device 120 determines whether dialogue request information, i.e., input information input by the user to the input / output terminal 110, is input from the input / output terminal 110 (step S401). The dialogue request information may be a signal transmitted when a start button is pressed by the user, or a signal transmitted when a button requesting settings or the like is pressed. The dialogue request information may include information about the person with whom the user wishes to dialogue, or may include information about the dialogue. The information about the dialogue may include the theme of the dialogue, the purpose of the dialogue, etc. The information about the person with whom the dialogue is to be dialogued may include at least one piece of information about the person, and may also include the number of people in the dialogue. The dialogue request information may include date and time information such as the date and day of the week.

[0038] If the request information for dialogue has not been input (No at step S401), the information processing device 120 returns to step S401 and waits for the input of the request information for dialogue.

[0039] When dialogue request information is input (step S401, Yes), the information processing device 120 analyzes the input dialogue request information and determines dialogue partner selection policy information based on the dialogue request information (step S402). The information processing device 120 may determine dialogue partner selection policy information using a table stored in the person DB 130.

[0040] The dialogue partner selection policy information is information for determining a dialogue partner generated by LLM 140, and may be information about at least one dialogue partner. That is, the dialogue partner selection policy information may be a person (name) or a person's characteristics. The dialogue partner selection policy information may be determined based on information about the user who inputs the dialogue request information, or may be determined based on the purpose of the dialogue. The dialogue partner selection policy information may be determined based on date and time information. For example, if the user plans to visit a tourist spot on the weekend, a person who is familiar with the tourist spot may be determined as the dialogue partner selection policy information. For example, the dialogue partner selection policy information may be determined based on the purpose of the dialogue determined based on date and time information. For example, if the user plans to give a presentation the next day, the purpose of the dialogue may be determined to be the presentation.

[0041] The compatibility between the user and the person stored in the person DB 130 may be determined by AI. For example, a learning model that learns compatible items for each item of information about the user and information about one or more people may be used as an input to determine information about compatible people. Compatibility may be determined by outputting the degree of compatibility for each item using an existing scoring model. Information about the person may be determined based on the compatibility. Each item of information about the user and the person with whom the conversation partner is based may be acquired from the person DB 130.

[0042] The selection policy information of the dialogue partner may be a person stored in the person DB 130. For example, if the user is characterized as being quiet, "Ahamo Taro," who is quite talkative, may be determined as the selection policy information of the dialogue partner.

[0043] Two or more people may be selected from person DB 130. The second person may be selected based on information about the first person. For example, selection policy information for the second person may be selected based on a table in which the compatibility between the user, the first person, and the second person is associated. The second person may be selected based on the characteristics of the user and the characteristics of the first person. For example, if the user is quiet and the first person is talkative, the second person selected from person DB 130 may be a person who is normally talkative.

[0044] The dialogue partner selection policy information may be information about a person other than the person stored in the person DB 130.

[0045] FIG. 5 shows an example of table 134 stored in dialogue DB 130 and used when determining dialogue partner selection policy information. FIG. 5 shows table 134 used when determining two people as dialogue partners. The number of dialogue partners does not have to be two, but may be one, or three or more. The information processing device 120 determines which user shown in FIG. 5 the user who inputs dialogue request information corresponds to, and determines the characteristics of the two people for each of the determined user's characteristics. Although FIG. 5 stores role and personality, other items such as value information and expertise may also be stored, and the user items and dialogue partner items do not have to be the same. If names are stored in table 134, step S403 can be omitted.

[0046] If the characteristics of the user who inputs the dialogue request information do not match the characteristics of the users shown in FIG. 5, a user who is closest to the user who inputs the dialogue request information is determined from among the users shown in FIG. 5, and the characteristics of the person corresponding to the determined user are determined. For example, if the user who inputs the dialogue request information is a talkative user, the information processing device 120 determines person 1 and person 2 corresponding to the top row as dialogue partner selection policy information. The number of people to be determined does not have to be two, but may be one, or may be three or more. The items listed in table 134 do not have to be roles and characteristics, but may be other items. For example, values ​​and expertise may be listed in table 134.

[0047] A plurality of tables 134 may be provided. The table 134 may be configured with a plurality of tables according to the purpose of the dialogue. For example, a table for a dialogue about a proposal with a colleague may be separate from a table for a rehearsal with a superior. The table 134 may be configured according to the user's emotions. The table 134 may be a table for each time period. For example, there may be separate tables for weekday daytime, weekday evening, and holidays.

[0048] The information processing device 120 determines a person based on the dialogue request information and the determined selection policy information (step S403). A person having the characteristics determined in step S402 is searched for in the person DB 130. The information processing device 120 may determine the name of the person. If a person satisfying all the characteristics is not present in the person DB 130, a person satisfying important characteristics is searched for first. For example, if role is more important than personality, a person matching the role is searched for first. Multiple people present in the person DB may be presented together with the degree of similarity so that the user can make a final decision (selection), or a person may be selected from multiple candidate people using a random number or a random number and the degree of similarity. Instead of selecting a person's name, the information processing device 120 may determine the characteristics of the person determined as the dialogue partner selection policy information as the dialogue partner, in which case step S403 can be omitted.

[0049] Based on the determined person, a prompt to be input to the LLM 140 is generated (step S404). When information about multiple people is input, the information processing device 120 may determine whether or not to use information about each person to generate a prompt based on feedback such as a user's decision. The information processing device 120 may determine whether or not to use information about each person to generate a prompt based on the number of times the "helpful" button is pressed. The information processing device 120 may determine that information about a person will be used to generate a prompt if the number of times the button is pressed is equal to or greater than a predetermined number, or may determine that information about a person will be used to generate a prompt if the button is pressed frequently.

[0050] 6 shows an example of a prompt 600 generated by the information processing device 120. The prompt 600 shows an example of a prompt when the LLM 140 is made to play the roles of Docomo Hanako, a female boss in her 30s, and Ahamo Taro, a male colleague in his 20s, and the LLM 140 is used to brush up an idea with Docomo Hanako and Ahamo Taro as conversation partners. The prompt 600 includes information 610 about the people and information 620 about the dialogue.

[0051] A person may be assigned multiple items in the person information 610. For example, a person may be assigned multiple roles. For example, three roles may be assigned: "to make the person aware of logical omissions and gaps," "to make suggestions that make the person aware of mistakes," and "to provide objective opinions on ideas presented in a dialogue." Furthermore, the items listed in the person information 610 may be items other than name, attributes, values, role, and talkativeness. When a person's name is listed as an item, an instruction may be given to retrieve other items related to the person with that name from the person DB 130.

[0052] The items in the person-related information 610 may be assigned a level of importance indicating how important each item is. For example, the level of importance may be "very important," "important," or "normal." The LLM 140 generates people taking the level of importance into consideration.

[0053] The information 620 about the dialogue may include reference sentences (example sentences) of what kind of statements should be made regarding the "role in the dialogue." For example, in the case of "making a suggestion that makes the customer realize the mistake," an example sentence such as "what would the customer think?" may be included. The example sentences may be stored in the person DB 130.

[0054] The information 620 about the interaction may include multiple purposes of the interaction, such as "brainstorm ideas together with the user" and "concreteize ideas."

[0055] By inputting the prompts shown in Figure 6 into the LLM 140, you can refine your ideas with your colleagues and also rehearse your proposal to your boss using the LLM 140 before submitting it to your boss for approval.

[0056] When the information processing device 120 inputs the generated prompt 600 to the LLM 140, it instructs the LLM 140 to start a conversation, and the conversation begins (step S405). The conversation may be, for example, a brush-up with a colleague on a new idea, or a rehearsal of an explanation to a superior. The user's utterances are input from the input device 122 and output to the LLM 140 via the output device 123. The utterances may be text or voice. The content of the dialogue is input to the input device 122 from the LLM 140 and output via the output device 123. The output may be voice, text, or both voice and text.

[0057] 7 shows an example of a conversation 700, which is the content of a dialogue using the LLM 140 after the prompt 600 is input. "User" is a utterance input by a user from the input / output terminal 110, and "Ahamo Taro" and "Docomo Hanako" are utterances of people generated by the LLM 140.

[0058] When a user said, "I thought of a service that would improve communication quality only when playing games," his colleague Ahamo Taro agreed, saying, "That's a good idea," and then expanded on the idea from a different perspective, saying, "It seems like it could be used for games, but maybe it could be used for videos too?"

[0059] In response to his colleague's suggestion, the user accepted the suggestion, saying, "It would also be good to use it for videos." He then expanded on the idea by saying, "It might be interesting if we could set the communication quality for each service to suit the "games" and "videos" communication quality."

[0060] In response to this idea, the user's boss, Docomo Hanako, made a comment to concretely flesh out the idea, asking, "Specifically, what level of communication quality should we set?" However, the user hadn't thought about it yet, so he was unable to answer his boss's question by simply saying, "Um..."

[0061] Before actually submitting documents to your superior for approval, you can practice by having a conversation like the one shown in Figure 7, and then create a document containing ideas that specify communication quality based on the conversation and propose it to your superior. This eliminates the need to take the proposed ideas home and later propose modifications that specify communication quality, saving time and the number of meetings.

[0062] <Use Case> The present disclosure can also be used for dialogues other than those conducted before preparing materials for approval by a superior. The dialogue may be role-playing for training or other developmental purposes, or dialogue in a role-playing game. The dialogue partner may be determined according to the difficulty level of the training or game. For example, when the difficulty level of the game is increased, a table associating people with poor compatibility may be used to select people with poor compatibility as dialogue partners, or information about people with poor compatibility may be determined using information about the user as input using a learning model that has learned poor compatibility items for each item of information about the user and information about one or more people. Compatibility may be determined by using an existing scoring model to output the degree of compatibility for each item. Information about people may be determined based on compatibility.

[0063] Furthermore, the dialogue can be used for brushing up ideas with colleagues, rehearsing for superiors, or dialogue other than role-playing. For example, when talking about characters from a favorite animation, the other characters from the same animation may talk about their favorite characters. For example, when deciding on a summer vacation destination, the destination may be decided by dialogue with a person who is knowledgeable about events held in various places and a person who is knowledgeable about local cuisine. For example, the social climate of the late Edo period may be discussed with Saigo Takamori, Sakamoto Ryoma, and Katsu Kaishu.

[0064] <Modification> The prompt 600 may be generated using historical information. For example, the information 610 about the person may include, for example, the content of statements made by the user in past conversations or emotions based on information obtained by analyzing the content of the statements. By setting emotions for the person generated by the user and / or the LLM 140 and engaging in a conversation, a more realistic conversation can be carried out. The emotions may increase the anger level and / or sadness level and decrease the happiness level when the user and / or the person makes an unconstructive comment.

[0065] In the above-described embodiments, the notation of each component may be replaced with other notations such as "...circuitry," "...assembly," "...device," "...unit," or "...module."

[0066] Furthermore, the present disclosure can be realized by software, hardware, or software linked to hardware. Each functional block used in the description of the above embodiments may be partially or entirely realized as an LSI, which is an integrated circuit, and each process described in the above embodiments may be partially or entirely controlled by a single LSI or a combination of LSIs. The LSI may be composed of individual chips, or may be composed of a single chip that includes some or all of the functional blocks. The LSI may have data input and output. Depending on the degree of integration, the LSI may be called an IC, system LSI, super LSI, or ultra LSI.

[0067] The integrated circuit method is not limited to LSI, but may be realized by a dedicated circuit, a general-purpose processor, or a dedicated processor. Also, a field programmable gate array (FPGA) that can be programmed after LSI manufacturing, or a reconfigurable processor that can reconfigure the connections and settings of circuit cells within the LSI, may be used. The present disclosure may be realized as digital processing or analog processing.

[0068] Furthermore, if an integrated circuit technology that can replace LSI emerges due to advances in semiconductor technology or other derivative technologies, it is natural that such technology can be used to integrate functional blocks. The application of biotechnology is also a possibility. [Industrial Applicability]

[0069] The present disclosure is useful for an information processing device that interacts with a user. [Explanation of symbols]

[0070] 100 systems 110 Input / Output Terminal 120 Information processing equipment 121 processors 122 Input Device 123 Output Device 124 Bus 130 Person DB 131 Person A's data 132 Person B's data 133 Person C's Data 134 Tables 140 LLM 600 prompt 610 Information about people 620 Dialogue Information 700 conversations

Claims

1. a processor that, when input information is input, determines selection policy information for determining a dialogue partner based on the input information, and generates a prompt based on the selection policy information; an output device connected to the generative AI model that generates the dialogue partner, the output device outputting the prompt; An information processing device comprising:

2. the processor generates the prompt by searching a knowledge database to determine the dialogue partner based on the selection policy information. The information processing device according to claim 1 .

3. The input information is dialogue request information including at least one of information about a user who inputs the input information, a purpose of the dialogue, and information about the dialogue partner. The information processing device according to claim 1 .

4. After outputting the prompt to the generative AI model, when a dialogue utterance is input, the processor outputs the utterance to the generative AI model. The information processing device according to claim 1 .

5. The information processing device Once the input information is entered, determining selection policy information for determining a dialogue partner based on the input information; generating a prompt based on the selection policy information; outputting the prompt to a generative AI model that generates the dialogue partner; Information processing methods.

6. On the computer, Once the input information is entered, determining selection policy information for determining a dialogue partner based on the input information; generating a prompt based on the selection policy information; outputting the prompt to a generative AI model that generates the dialogue partner; A program for performing information processing.

Images