Program, information processing method, and information processing apparatus

Abstract

To provide a program and the like capable of specifying a problem of a user and proposing an action to be performed for the specified problem.SOLUTION: A computer specifies a problem of a user on the basis of an interaction with the user according to a program. Further, the computer determines a plurality of actions to be performed for the specified problem, and estimates an evaluation in a case where each determined action is performed. Then, the computer outputs each determined action and an evaluation when each action is performed.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a program, an information processing method, and an information processing device. [Background technology]

[0002] When doing something, users find a problem and move closer to achieving it through repeated trial and error. At this time, it is important to set an appropriate problem and determine the correct (appropriate) approach to solving the problem. However, it is not easy to determine whether the problem is appropriate, whether the approach to solving the problem is appropriate, or whether the approach being implemented is bringing the problem closer to being solved. Patent Document 1 discloses a technology that can grasp the relationship (correlation or causation) between three or more variables. According to the technology disclosed in Patent Document 1, various types of data (multivariate data) are collected, and when a problem (problem) arises, correlation analysis or causal analysis can be applied to the collected multivariate data to identify the factors that contributed to the problem. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] International Publication No. 2020 / 004154 Summary of the Invention [Problem to be solved by the invention]

[0004] However, while the technology disclosed in Patent Document 1 can identify the factors that caused an issue (problem) when it occurs, it does not extract the issues that the user needs to solve, and it is difficult to suggest the issues that the user needs to solve and the actions that the user should take in response to the issues.

[0005] The present disclosure aims to provide a program or the like that is capable of identifying a user's problem and proposing an action to be taken for the identified problem. [Means for solving the problem]

[0006] A program according to one aspect of the present disclosure causes a computer to execute a process of identifying a user's issues based on a dialogue with the user, determining multiple actions to be taken for the identified issues, estimating an evaluation for each of the determined actions, and outputting each of the determined actions and an evaluation for each of the determined actions. [Effects of the Invention]

[0007] According to the present disclosure, it is possible to identify a user's problem and propose an action to be taken for the identified problem. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a block diagram illustrating an example of the configuration of an information processing device. [Figure 2] FIG. 2 is an explanatory diagram illustrating an example of the configuration of a DB included in a knowledge base. [Figure 3] FIG. 2 is an explanatory diagram of a process performed by an information processing device. [Figure 4] 10 is a flowchart illustrating an example of a procedure for presenting an action. [Figure 5] FIG. 10 is an explanatory diagram showing an example of a screen. [Figure 6] FIG. 10 is an explanatory diagram showing an example of a screen. [Figure 7] 10 is a flowchart illustrating an example of a procedure for presenting an evaluation of an action. [Figure 8] FIG. 10 is an explanatory diagram showing an example of a screen. [Figure 9] FIG. 10 is an explanatory diagram showing an example of a screen. [Figure 10] FIG. 10 is an explanatory diagram showing an example of a screen. [Figure 11] FIG. 10 is an explanatory diagram showing an example of a screen. [Figure 12]FIG. 10 is an explanatory diagram of a feedback process of evaluations from users. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, a program, an information processing method, and an information processing device according to the present disclosure will be described in detail with reference to the drawings illustrating embodiments thereof.

[0010] In this embodiment, an information processing device is described that uses general-purpose large-scale language models (LLMs) to engage in a dialogue with a user, extracts the user's issues based on the dialogue content and a knowledge database that stores information about the user's industry and business, and presents actions to solve the extracted issues. In this embodiment, an example is described in which Japanese is used as the language input to the large-scale language model, but other languages ​​may also be used.

[0011] FIG. 1 is a block diagram showing an example of the configuration of an information processing device. The information processing device 10 is a computer capable of various information processing and transmitting and receiving information, such as a server computer, a personal computer, or a tablet terminal. The information processing device 10 includes a control unit 11, a storage unit 12, a communication unit 13, an input unit 14, a display unit 15, a reading unit 16, and the like, which are interconnected via a bus. The control unit 11 is configured using one or more processors, such as a central processing unit (CPU), a micro-processing unit (MPU), a graphics processing unit (GPU), a general-purpose computing on graphics processing units (GPGPU), a tensor processing unit (TPU), or an AI chip (semiconductor for AI). The control unit 11 executes various information processing and control processes to be performed by the information processing device 10 by appropriately executing a program P stored in the storage unit 12. Note that if the control unit 11 includes multiple processors, the control unit 11 may execute each process (each function) using a different processor.

[0012] The storage unit 12 includes a RAM (Random Access Memory), a flash memory, a hard disk, an SSD (Solid State Drive), etc. The storage unit 12 stores a program P (program product) executed by the control unit 11 and various data required for executing the program P. The storage unit 12 also temporarily stores data, etc., generated when the control unit 11 executes the program P. The storage unit 12 also stores a language model M. The language model M is a general-purpose large-scale language model constructed by pre-training using a large set of sentences, etc. The language model M is trained so that, when input data including, for example, text data written in a natural language is input, the language model M performs a calculation to generate output data corresponding to the content of the input data based on the input data and outputs the generated output data. The language model M can be configured using an algorithm such as GPT (Generative Pre-trained Transformer)-3, GPT-3.5, or GPT-4, or may be configured by combining multiple algorithms. Note that the language model M is not limited to the Transformer-based model described above. The language model M is expected to be used as a program module constituting artificial intelligence software. The language model M performs a predetermined calculation on input data and outputs the calculation result, and data such as coefficients and thresholds of functions that define this calculation are stored as the language model M in the storage unit 12. Instead of storing the language model M in the storage unit 12, the information processing device 10 may access a language processing server that stores the language model M and read it out. Furthermore, the storage unit 12 stores a knowledge base 12a (knowledge database). The storage unit 12 may be composed of multiple storage devices, and part of the storage unit 12 may be another storage device connected to the information processing device 10 or another storage device with which the information processing device 10 can communicate.

[0013] The communication unit 13 is a communication module for performing processes related to wired or wireless communication, and transmits and receives information to and from other devices via a network. The network may be the Internet or a public telephone network, or a LAN (Local Area Network) established within the facility where the information processing device 10 is installed. The input unit 14 accepts operation input by a user and sends a control signal corresponding to the operation content to the control unit 11. The input unit 14 includes, for example, a keyboard, a mouse, and a microphone for voice input. The display unit 15 is a liquid crystal display or an organic EL (electroluminescence) display, and displays various information according to instructions from the control unit 11. A part of the input unit 14 and the display unit 15 may be a touch panel integrally configured.

[0014] The reading unit 16 reads information stored in a portable storage medium 10a, which may include a CD (Compact Disc), a DVD (Digital Versatile Disc), a USB (Universal Serial Bus) memory, an SD (Secure Digital) card, etc. The program P and data stored in the storage unit 12 may be read by the control unit 11 from the portable storage medium 10a via the reading unit 16 and stored in the storage unit 12. The program P and data may also be written to the storage unit 12 during the manufacturing stage of the information processing device 10, or may be downloaded by the control unit 11 from another device via the communication unit 13 and stored in the storage unit 12.

[0015] In this embodiment, the information processing device 10 may be a multi-computer consisting of multiple computers, or may be a virtual machine virtually constructed by software within a single device. Furthermore, when the information processing device 10 is configured as a server computer, the information processing device 10 may be a local server installed in the facility where the information processing device 10 is located, or a cloud server connected to the information processing device 10 via a network such as the Internet. The following description assumes that the information processing device 10 is a single computer. Furthermore, the program P may be deployed on a single computer or a single site, or may be distributed across multiple sites and executed on multiple computers interconnected via a network. Furthermore, the information processing device 10 does not necessarily require the input unit 14 and the display unit 15. It may also be configured to accept operations via a connected computer or to output information to be displayed to an external display device. The information processing device 10 may also be configured to include a speaker for audio output.

[0016] The information processing device 10 having the above-described configuration engages in a dialogue (question and answer session) with a user, extracts the user's issues from the dialogue, and identifies and presents multiple goals for solving the extracted issues. The user determines and selects a goal they wish to achieve from the multiple presented goals. The information processing device 10 accepts the goal selected by the user and identifies and presents multiple actions the user can take to achieve the selected goal. The user selects an action they wish to execute from the multiple presented actions and executes the selected action. After executing the action, the user evaluates whether or not the action has brought them closer to achieving the selected goal and whether or not they have brought them closer to solving the original issue. The information processing device 10 accepts the action selected (executed) by the user and the user's evaluation after executing the action, and generates and presents a degree of achievement for solving the original issue at this point in time. The information processing device 10 also identifies and presents the next action the user can take to solve the original issue and achieve the user's selected goal based on the degree of achievement at this point in time, etc. Alternatively, the information processing device 10 identifies and presents another goal (next goal) for solving the original problem based on the current level of achievement, etc. The user determines the goal they want to achieve from the goals presented sequentially, selects and executes the action they want to perform from the actions presented sequentially, and provides feedback on the evaluation of the execution result to the information processing device 10. Each time the user performs an action, the information processing device 10 generates and presents the level of achievement at each point in time based on the user's evaluation after performing the action, and identifies and presents the next action or next goal that the user can take. By repeating such processing between the information processing device 10 and the user, the user can get closer to solving the problem extracted from the initial dialogue content. The information processing device 10 of this embodiment, which performs the above-mentioned processing, uses the language model M and the knowledge base 12a when interacting with the user, extracting a problem from the dialogue content with the user, identifying a goal for solving the problem, identifying an action for achieving the goal, and generating a level of achievement based on the user's evaluation after performing the action.

[0017] The language model M may be a general-purpose large-scale language model that is used as is, or may be fine-tuned using information on the business type, industry, occupation, business, etc. of the user's company. The language model M may be fine-tuned by the information processing device 10 or by another learning device. The language model M fine-tuned by the other learning device is downloaded from the learning device to the information processing device 10 via a network or a portable storage medium 10a, for example, and stored in the storage unit 12.

[0018] FIG. 2 is an explanatory diagram showing an example of the configuration of a database included in the knowledge base 12a. The knowledge base 12a (knowledge database) includes an action database and a similar action group database shown in FIG. 2. The action database is a database that stores history information about actions actually performed by a user among actions presented to the user by the information processing device 10. The action database stores, in association with each other, the identification information (action ID) assigned to each action, the identification information (user ID) of the user who selected each action, a timestamp indicating the date and time when the user selected each action, the identification information (task ID) of the user's assignment, the identification information (similar action group ID) of the similar action group to which each action belongs, the content of each action, related information, other actions (action IDs) presented together with each action, the action vector of each action, and the feedback method (feedback type) provided by the user who performed (selected) each action. The related information is information about companies, technologies, etc. related to the action. The action vector is a vector representation generated from the attributes of each action and is used to measure the similarity with other actions.

[0019] Actions registered in the action DB are grouped into similar actions based on similarity based on action vectors. The similar action group DB stores the number of selections and the like for each action grouped into each similar action group in association with a similar action group ID. The number of selections stored in the similar action group DB is used as a factor in determining whether or not to select an action to be presented to the user. In this embodiment, when a user selects an action from the actions presented to the user by the information processing device 10, 1 is added to the number of selections of the similar action group to which the selected action belongs, and the more the number of selections, the more likely the action will be selected as an action to be presented to the user in subsequent processing.

[0020] The knowledge base 12a contains various types of information generated using, for example, the technology of the Causal Information Analysis Service (CALC) developed by Sony Computer Science Laboratories, Inc. The knowledge base 12a includes various types of information (knowledge), such as information (knowledge) collected from dialogues with users, public information collected via a network, closed knowledge with limited disclosure scope (non-public information, confidential information, etc.), and information acquired by the information processing device 10 from past performance. Public information includes knowledge learned by an LLM (Language Model M) such as ChatGPT (registered trademark), information published on the Web, and information (knowledge) obtained by scraping this information, including corporate information, social conditions, market information, population information, legal regulations, systems, and standardization information. Closed knowledge includes research findings and various data currently under research. Closed knowledge also includes information collected from dialogues with users, such as information on the user's environment, information on the user's or company's resources, the user's awareness of issues, concerns, actual challenges, and goals. Furthermore, the closed knowledge includes information such as user evaluation results (such as past case performance) fed back for each action. Each piece of information (knowledge) included in the knowledge base 12a may include data other than text data, such as table data, numerical data, and image data. The information processing device 10 determines action candidates based on the various types of information described above, and identifies the relationships (correlation and causation) between each action through correlation analysis and causal analysis (causal inference), thereby generating the knowledge base 12a that retains the attributes of each action and the relationships (correlation and causation) between each action. In addition, the knowledge base 12a receives feedback of evaluations after users perform the presented actions, and thereby successively updates the attributes and relationships of each action.

[0021] Next, each function realized by the control unit 11 will be described. The control unit 11 realizes each function of an initial value setting unit 111, a problem decomposition unit 112, an action presentation unit 113, a feedback acquisition unit 114, an evaluation confirmation unit 115, and an action evaluation learning unit 116 by appropriately executing a program P stored in the storage unit 12. The initial value setting unit 111 engages in a dialogue with the user and collects the user's awareness of problems, worries, actual problems, goals, etc. from the dialogue content. Specifically, the initial value setting unit 111 generates a question message that elicits the user's awareness of problems, worries, actual problems, goals, etc., displays it on the display unit 15, and asks the user a question, and acquires the user's answer to the question via the input unit 14. Note that the question message to be asked initially may be stored in advance in, for example, the storage unit 12. The initial value setting unit 111 may also output the question message by voice. The initial value setting unit 111 acquires the user's answer to a question as text data written in a natural language, such as text data input via the input unit 14 or text data converted from voice data acquired by voice input of the user's voice uttered in a natural language. When engaging in a dialogue with the user, the initial value setting unit 111 generates response information (e.g., a further question message) to a message from the user (e.g., the user's answer to a question) using the language model M. The initial value setting unit 111 also collects, from the content of the dialogue with the user, information on the environment in which the user is placed (e.g., company management figures, factory production volume, health status of each employee, past performance and achievements of the company or user, etc.), information on resources possessed by the user or company (e.g., human resources, equipment, experience, funds, networks with other companies, etc.), information on the market environment in which the user and company belong, etc. (hereinafter, these may be collectively referred to as user environment information). In addition to the content of the dialogue with the user, the initial value setting unit 111 may collect various types of information that are publicly available via a network such as the Internet, or may cooperate with a system used by the user or the user's company to automatically collect various types of information from the system. Furthermore, the initial value setting unit 111 may acquire information on problems that the user has solved in the past and goals that the user has already achieved as prerequisites.

[0022] The problem decomposition unit 112 extracts the user's problems from the various information collected by the initial value setting unit 111 and breaks down the extracted problems into multiple goals for solving the problems and actions the user can take to achieve each goal. The problem decomposition unit 112 analyzes the information collected by the initial value setting unit 111 by referring to information stored in the knowledge base 12a to extract problems, and extracts goals for solving the problems and actions for achieving each goal. For example, the problem decomposition unit 112 performs natural language processing on sentences obtained from the content of a dialogue with the user to analyze their meaning, analyzes the user's situation based on numerical values ​​and other information obtained from market environment information, and automatically extracts problems from the analysis results. Note that if the initial value setting unit 111 collects information on past cases, the problem decomposition unit 112 also determines the type of problem by comparing it with past cases so that appropriate actions can be suggested. The type of problem can be, for example, "problem related to factory productivity" or "problem related to cost reduction." Furthermore, when the initial value setting unit 111 collects prerequisites, the problem decomposition unit 112 excludes problems that have already been solved based on the prerequisites from the processing targets (presentation targets). Furthermore, if a previously presented action has received a negative evaluation, the problem decomposition unit 112 may exclude that action from the presentation targets. Furthermore, even if a previously presented action has received a negative evaluation, the problem decomposition unit 112 may exclude that action from the presentation targets in subsequent processing if it is presented again and receives a negative evaluation again, or if it receives negative evaluations a predetermined number of times or more. Furthermore, if the problem decomposition unit 112 cannot automatically extract a problem, the initial value setting unit 111 may again interact with the user to extract further information and extract the problem. Furthermore, even if the user is aware of their own problem, the problem decomposition unit 112 may extract the problem by referring to information in the knowledge base 12a. In this case, it is possible to present a problem that is more suitable for the user's situation than the problem the user is aware of.

[0023] The action presenter 113 presents the goals and actions generated (extracted) by the problem decomposition unit 112 to the user, for example, by displaying them on the display unit 15. Specifically, the action presenter 113 first presents multiple goals, accepts the user's selection of a goal they wish to achieve, and then presents multiple actions for achieving the selected goal. When presenting actions, the action presenter 113 may estimate the effect of each action (evaluation score and achievement level for the task if each action is performed) using, for example, a simulation process, and present the estimated effect along with the action content. Furthermore, the action presenter 113 may determine a recommendation level for each of multiple actions for each goal based on criteria such as the possibility of achieving the task (achievement level), the challenge level, and the risk level, and present each action in descending order of recommendation level. The actions may be presented in descending order of the newest services or technologies that realize the actions, or in descending order of popularity.

[0024] The user decides (selects) an action to be executed (implemented) from multiple actions presented by the action presenting unit 113, and the action presenting unit 113 accepts the user's selection of the action to be executed. The action presenting unit 113 sends the action selected by the user to the action evaluation learning unit 116. The action evaluation learning unit 116 sends information about the user's action selection (the event that the user has selected the action) to the problem decomposition unit 112 as re-learning data, and the problem decomposition unit 112 re-learns information to be used when extracting actions in the future. For example, the problem decomposition unit 112 updates the relationship between the action selected by the user and other actions that have a relationship with this action in the knowledge base 12a, and re-learns (updates) the criteria for extracting actions by referring to the knowledge base 12a.

[0025] If the user is not satisfied with the presented action, he or she can request the information processing device 10 to present a different action, or can start over by having the problem decomposition unit 112 extract the problem. In this case, the action presenting unit 113 sends an event that the user did not select the presented action to the action evaluation learning unit 116 as user selection information for the action. The action evaluation learning unit 116 sends this user selection information to the problem decomposition unit 112 as re-learning data, and performs re-learning so that the problem decomposition unit 112 will exclude the action from extraction targets when extracting actions in the future, or will make corrections to increase the likelihood that the user will select the action.

[0026] After executing the selected action, the user evaluates the executed action and feeds back the evaluation result (evaluation information) to the information processing device 10. For example, the user evaluates whether performing the action has brought the user closer to solving the original problem (the problem extracted by the problem decomposition unit 112) or the selected goal (the goal selected by the user from the goals generated by the problem decomposition unit 112). The feedback acquisition unit 114 receives the evaluation information feedback via an input screen for receiving the user's input of the evaluation information. The input screen has a different configuration depending on the content of the action selected by the user, and may include parameters of the effect presented together with the action content when the action is presented. For example, the input screen may have an input field for an evaluation value expressed on a five-point scale, an input field for a numerical value for items that can be quantitatively evaluated, and an input field for a free word for items that can be qualitatively evaluated. The user inputs the evaluation information for each item into each input field of the input screen via the input unit 14.

[0027] The evaluation confirmation unit 115 acquires the evaluation information from the user acquired by the feedback acquisition unit 114, and evaluates the action performed by the user, taking into consideration past case performance and the like. The evaluation confirmation unit 115 also updates various information (e.g., user environment information) collected by the initial value setting unit 111 to information after the action is performed, acquires the degree of change, and evaluates the action performed by the user based on the degree of change. The evaluation confirmation unit 115 evaluates the action by performing processing according to the type of evaluation information from the user and the type of information acquired for the degree of change. The evaluation confirmation unit 115 has a processing unit and a set of evaluation functions (both not shown), and the processing unit and the set of evaluation functions are prepared according to the type of evaluation information from the user and the type of information acquired for the degree of change. When the evaluation information from the user and the degree of change include text data written in natural language, the processing unit performs a positive / negative determination process on the text data using natural language processing, and inputs the result into the evaluation function. The evaluation function here is an evaluation function for determining whether the action has been effective or not based on the positive / negative judgment result for the text data. Therefore, the evaluation confirmation unit 115 inputs the positive / negative judgment result for the text data into the evaluation function here, and acquires a KPI (Key Performance Indicator) score that indicates the degree of achievement of the original task (the task extracted by the task decomposition unit 112) by the action.

[0028] Furthermore, if the user's evaluation information and degree of change include numerical data such as accounting data (e.g., sales, costs, etc.), the processing unit calculates the amount of change (increase or decrease) from the past (e.g., most recent) numerical data and inputs the calculation result into the evaluation function. The evaluation function here is an evaluation function for determining whether the action has been effective based on the amount of change in the numerical data. Therefore, the evaluation confirmation unit 115 inputs the amount of change in the numerical data into the evaluation function here and obtains a KPI score indicating the degree of achievement of the original task through the action. Furthermore, if the user's evaluation information includes an evaluation value scored by the user (e.g., an evaluation value expressed on a five-point scale), the processing unit inputs the evaluation value scored by the user into the evaluation function. Note that the processing unit may input past evaluation values ​​(e.g., the most recent or recent few times) into the evaluation function in addition to the evaluation value scored by the user here. The evaluation function here is an evaluation function for determining whether the action has been effective based on the user's evaluation value. Therefore, the evaluation confirmation unit 115 inputs the user's evaluation value into this evaluation function and obtains a KPI score that indicates the degree of achievement of the original task by the action. The evaluation function group is not limited to the above-mentioned evaluation functions, and may include, for example, an evaluation function for determining the effect of the action from time changes in values ​​detected by various sensors (sensor values). The evaluation confirmation unit 115 calculates an overall KPI score for the action from the KPI score calculated using the evaluation function from the evaluation information from the user and the degree of change in various information collected by the initial value setting unit 111.

[0029] When the evaluation confirmation unit 115 calculates the KPI score at this point, it plots it on a graph with the elapsed time on the horizontal axis and the KPI score on the vertical axis to generate a difference product graph showing the degree of KPI integration over time. The difference product graph is, for example, the graph shown as G1 in FIG. 8, G2 in FIG. 9, and G3 in FIG. 10, and the effect of each action is indicated by an arrow representing the start and end times of each action and the slope corresponding to the amount of change in the KPI score due to each action. The greater the slope of the difference product graph, the greater the effect of the action (positive), and the smaller the slope of the difference product graph, the smaller the effect of the action (negative). Note that the content of the KPI score indicated by the vertical axis of the difference product graph differs depending on the type of issue. For example, if the type of issue is a "factory productivity issue," it shows the degree of effect on improving factory productivity, and if the type of issue is a "cost reduction issue," it shows the degree of effect on cost reduction. The evaluation confirmation unit 115 can present the generated difference product graph on the display unit 15 so that the user can confirm the effect of the action. The feedback acquisition unit 114 acquires evaluation information from the user after the user executes the action, and periodically in the case of an action that is executed continuously. Therefore, the evaluation confirmation unit 115 can evaluate the action being executed at the timing when the feedback acquisition unit 114 acquires evaluation information from the user.

[0030] The action evaluation learning unit 116 determines whether the effect of the action is evaluated positively or negatively based on the KPI score calculated by the evaluation confirmation unit 115, and sends the evaluation result of the effect of the action to the problem decomposition unit 112 as re-learning data, so that the problem decomposition unit 112 re-learns information to be used when extracting actions in the future. For example, the problem decomposition unit 112 updates the relationship between each action and other actions in the knowledge base 12a according to the evaluation result, and re-learns (updates) the judgment criteria for extracting actions by referring to the knowledge base 12a. For example, the action evaluation learning unit 116 adds 1 to the number of times a similar action group to which a positively evaluated action belongs in the similar action group DB included in the knowledge base 12a has been selected. The action evaluation learning unit 116 may also subtract 1 or a predetermined number from the number of times a similar action group to which a negatively evaluated action belongs, or may set the number of times the similar action group has been selected to zero. In addition, the action evaluation learning unit 116 may be configured to calculate the degree of positive evaluation or negative evaluation of the effect of the action from the KPI score calculated by the evaluation confirmation unit 115, and the problem decomposition unit 112 may perform re-learning depending on the calculation result.

[0031] FIG. 3 is an explanatory diagram of the processing performed by the information processing device 10. Using the above-described functions, the control unit 11 of the information processing device 10 extracts the user's problem from the content of the dialogue with the user, breaks down the extracted problem into multiple goals for solving the problem, and extracts multiple actions the user can take to achieve each goal. The user selects one goal they want to achieve from the multiple goals, and then selects one action they want to execute from multiple actions for achieving the selected goal. The goals and actions indicated by hatching in FIG. 3 indicate the goal and action selected by the user. After executing the selected action, the user evaluates the effect obtained as a result of the execution and provides feedback to the information processing device 10. The control unit 11 generates and evaluates the degree of achievement (KPI score) of the solution of the original problem at this point in time based on the user's evaluation and information in the knowledge base 12a. This allows the control unit 11 to obtain an evaluation (KPI score) of the action selected and executed by the user.

[0032] Next, the control unit 11 extracts and presents the next action to be taken by the user to achieve the goal selected by the user, based on the evaluation of the action performed by the user and the information in the knowledge base 12a (see arrow A in FIG. 3). The next action to be taken may include the action selected and performed by the user. The user selects one action to be performed from the presented actions, performs the selected action, and feeds back an evaluation of the execution result to the information processing device 10. The control unit 11 then evaluates the degree of achievement (KPI score) for solving the original problem at this point. In this way, after evaluating the performed action, the control unit 11 repeatedly presents the next action to be taken by the user, so that the user can select and perform the desired action from the actions presented in sequence, thereby getting closer to solving the original problem.

[0033] The control unit 11 may also extract and present a next goal for solving the original problem and an action for achieving the next goal based on the evaluation of the action performed by the user and information in the knowledge base 12a (see arrow B in FIG. 3). The user selects one goal to achieve from the presented goals, and selects one action to execute from multiple actions for achieving the selected goal. The user then executes the selected action and provides feedback on the evaluation of the execution result to the information processing device 10, and the control unit 11 evaluates the degree of achievement of the original problem at this point. In this way, the control unit 11 repeatedly presents the next goal for solving the original problem and the next action the user should take to achieve this goal, and the user can approach the solution of the original problem by selecting and executing the desired goal and action from the goals and actions presented sequentially.

[0034] Furthermore, the control unit 11 may extract and present the next task based on the evaluation of the action performed by the user and information in the knowledge base 12a (see arrow C in FIG. 3). In this case, the control unit 11 extracts and presents from the new task multiple goals for solving the task and multiple actions the user can take to achieve each goal. This allows the user to select one goal they want to achieve from the multiple goals and select and execute one action they want to perform from multiple actions for achieving the selected goal. The user also provides feedback on the evaluation of the results of the action performed to the information processing device 10, and the control unit 11 evaluates the degree of achievement of the new task at this point. In this way, the control unit 11 may change (update) the task based on the evaluation of the user's actions, and repeatedly presents the goals for solving the task and the actions for achieving each goal for the updated task. This allows the user to sequentially solve tasks that change over time and reach their desired goal.

[0035] After evaluating the degree of achievement at the time when the user performed an action, the control unit 11 may determine whether to present the next action, the next goal, or a new task based on the degree of achievement at that time, or based on the content of the dialogue with the user at that time, or based on a selection by the user. Furthermore, when the control unit 11 extracts the next action to be presented, the control unit 11 may be configured to present the next goal or the next task if the degree of achievement according to the simulation results of each extracted action is low.

[0036] The following describes a process in which the information processing device 10 of this embodiment interacts with a user, extracts the user's issues from the content of the interaction, and presents actions to be taken for the extracted issues. Fig. 4 is a flowchart showing an example of the action presentation process, and Figs. 5 and 6 are explanatory diagrams showing example screens. The control unit 11 of the information processing device 10 executes the following process in accordance with a program P stored in the storage unit 12.

[0037] When a user starts up the information processing device 10, the control unit 11 (initial value setting unit 111) of the information processing device 10 displays an interactive screen as shown in FIG. 5A on the display unit 15 (S11). The screen of FIG. 5A displays a question message, "Is there anything you would like to improve?" as the first question. The screen of FIG. 5A is provided with an input field R1 for the user to input a message, and the user inputs an answer to the displayed question in the input field R1 and operates the send button. The initial value setting unit 111 determines whether or not input of a message (an answer to the first question) from the user has been accepted via the input field R1 (S12), and if it determines that the message has not been accepted (S12: NO), it waits until it is accepted.

[0038] When the control unit 11 (problem decomposition unit 112) determines that it has received an input of an answer message from the user (S12: YES), it extracts keywords that represent the issues the user is facing from the answer message (S13). For example, the problem decomposition unit 112 adds a message to the answer message from the user saying, "Please extract keywords that represent the issues from this message," to create a prompt, and inputs this into the language model M, and obtains the keywords (issues) extracted from the answer message from the user using the language model M. In the example of FIG. 5B, the user inputs, "I want to improve the cost of disposing of coffee grounds," and the problem decomposition unit 112 extracts, for example, the issues of "coffee grounds disposal" and "cost improvement."

[0039] The control unit 11 (initial value setting unit 111, problem decomposition unit 112) generates response information such as a further question to be asked to the user based on the problem extracted from the answer message from the user, or a goal decomposed from the problem based on the answer message from the user (S14), and displays the generated response information on the display unit 15 (S15). The control unit 11 obtains the further question or goal for solving the problem generated by the language model M by creating a prompt such as "Based on the content of the answer from the user, please extract a goal for solving the problem. If you cannot extract a goal, please ask the user further questions that will enable you to extract the goal" and inputting this into the language model M.

[0040] The control unit 11 determines whether the displayed response information is a question for the user (S16). If it determines that the displayed response information is a question for the user (S16: YES), the process returns to step S12 and continues accepting input of response messages from the user in the input field R1. Each time a message is accepted from the user via the input field R1, the problem decomposition unit 112 extracts the user's problem based on all the messages entered (S13) and generates response information for the user (S14). In the example of FIG. 5B, the control unit 11 asks the user three questions and obtains answers to each question from the user. Note that the problem decomposition unit 112 does not need to extract a problem from the message each time it accepts a message from the user; once a problem is extracted, the process of step S13 may be skipped.

[0041] If the control unit 11 (action presenting unit 113) determines that the displayed response information is not a question for the user (S16: NO), i.e., if the displayed response information is a goal generated based on the user's answer message, the control unit 11 accepts the selection of one of the displayed goals via the input field R1 (S17). In the example of FIG. 5B, the message M1 is the response information presenting multiple goals, and the user responds to this message by inputting an answer message M2 indicating that the user has selected the second presented goal. If the control unit 11 (problem decomposition unit 112) accepts the answer message indicating the selection of one of the goals, the control unit 11 extracts actions for achieving the selected goal and generates action information (S18). Here, the control unit 11 generates action information by acquiring actions extracted (generated) by the language model M, for example, by creating a prompt such as "Please extract specific actions for methods of reusing coffee grounds other than for animal feed" and inputting it into the language model M. The control unit 11 may also extract actions for achieving the selected goal by referring to information (such as information on past cases) stored in the knowledge base 12a. Specifically, the control unit 11 extracts information on past cases that had the same problem as the current problem from the knowledge base 12a, creates a prompt that instructs the generation of an action based on the extracted information and the current problem, and inputs the prompt to the language model M, thereby obtaining the action information generated by the language model M. Here, the control unit 11 may also re-engage in a dialogue with the user, create a prompt based on the dialogue content and the knowledge base 12a, and obtain the action information generated by the language model M.

[0042] The control unit 11 (action presentation unit 113) displays the generated action information on the display unit 15 (S19). In the example of FIG. 5B, as shown by M3, information on four actions extracted by the problem decomposition unit 112 as methods for reusing coffee grounds is presented. When the problem decomposition unit 112 extracts actions for achieving the goal selected by the user, it simulates the execution of each action using a simulator based on the user's environmental information to measure the feasibility of achieving the task (degree of achievement), the degree of challenge (difficulty), the degree of risk, etc., and determines a recommendation level based on the values ​​of these indicators and presents the recommendation level along with the action content. The recommendation level may be, for example, the average value of each indicator. Furthermore, if there is a sufficient track record of presenting the same action in the past, the problem decomposition unit 112 may calculate the recommendation level by taking into account the ratio of the number of cases in which the presented action was selected to the total number of presented cases, or the ratio of cases in which the presented action was selected and the user gave a positive evaluation after the action was performed. In the example of FIG. 5B, the action presentation unit 113 presents each action in descending order of recommendation level. Furthermore, the problem decomposition unit 112 may simulate the execution of each extracted action using a simulator and present the simulation results together with the content of the action. Fig. 6 shows another example of actions presented by the action presentation unit 113. In the example of Fig. 6, the content of the action, the recommendation level of the action, and the results of simulating the execution of the action are displayed for each action.

[0043] The problem decomposition unit 112 may also calculate the profits expected from each action based on the simulation results and generate and present a chart displaying the expected profits from each action, as shown by the bar graph in FIG. 6. The bar graph in the upper right of FIG. 6 shows that, while previously, a cost for disposing of coffee grounds was required, the proposed action will result in profits from the coffee grounds. Simulation results may also be presented as text data, or, for example, when multiple companies are compared, a bar graph showing the profits expected from adopting each company. Furthermore, the problem decomposition unit 112 may estimate the degree of achievement (KPI score, evaluation score) of the original task if each action is taken based on the results of simulating the execution of each action, and generate and present a chart showing the change in the degree of achievement of the original task over time using arrows, as shown in the graph in the lower right of FIG. 6. Such a chart allows the effects of each proposed action to be easily compared based on the slope of each arrow. In addition, the problem decomposition unit 112 may obtain the KPI score generated in the language model M by creating a prompt including the simulation results or the KPI score estimated from the simulation results, and evaluation information by the user, and including a message such as "Please estimate the KPI score that indicates the degree of achievement for the original task," and inputting this into the language model M.

[0044] In addition to being configured to perform simulations, the problem decomposition unit 112 may be configured to estimate physical property information (physical property values) of an object of an action, estimate the failure rate of the object, etc., using a learning model constructed by machine learning. The problem decomposition unit 112 may also use a learning model to estimate people's emotions (sensitivity) toward the object, reactions on social media, etc. For example, a learning model configured using algorithms such as CNN (Convolution Neural Network) or Transformer, which is trained to output information about the object's physical property information, failure rate, emotions, reactions on social media, etc., when information about the object of an action is input, can be used. In this case, information about the object obtained by each action can be presented along with the content of each action.

[0045] Through the above-described process, the information processing device 10 performs a simulation for each action to be presented to the user, presents the results to the user, and the user decides which action to actually execute based on the simulation results. The effect of each action can only be evaluated by actually executing it, but by estimating the effect of each action through simulation, it is possible to obtain the results of virtually executing multiple actions in parallel, thereby virtually realizing a parallel world. By allowing the user to select an action that will produce optimal results based on the execution results of each action virtually executed in the parallel world, it is expected that the increase in the number of trial and error steps when developing a business, product, etc. will be suppressed, contributing to the promotion of development.

[0046] The control unit 11 (action presenting unit 113) accepts the selection of one of the displayed actions via the input field R1 (S20). In the example of FIG. 5B, as shown by M4, a reply message notifying that the first presented action has been selected is input. The control unit 11 (action evaluation learning unit 116) performs re-learning using the user's action selection information (the event that the user selected one of the actions) as re-learning data (S21). Specifically, the control unit 11 adds 1 to the number of selections of the similar action group to which the selected action belongs in the similar action group DB of the knowledge base 12a. The control unit 11 (action evaluation learning unit 116) may also perform re-learning using the user's target selection information (the event that the user selected one of the targets) accepted in step S17. This updates the information held in the knowledge base 12a and also updates the criteria for extracting actions by referring to the knowledge base 12a.

[0047] The above-described process uses knowledge base 12a, generated from information on past cases, to extract the user's issues from the dialogue with the user and present actions to the user. The user selects an action to execute from the presented actions, and the event for which the user selected the action is updated in knowledge base 12a. This allows the user to update the criteria for determining whether or not to extract the action to present to the user in subsequent processes. Note that in the above-described process, if the user is not satisfied with the presented action, they can request a different action via input field R1 or start over from problem decomposition. In this case, the presented action may be negatively evaluated as an action not selected by the user, and the negative evaluation of this action may also be updated in knowledge base 12a.

[0048] Through the above-described process, a user who selects one of the actions presented by the information processing device 10 executes the selected action. After executing the action, the user evaluates the action and feeds back the evaluation information to the information processing device 10. When the information processing device 10 receives feedback of the user's evaluation, it calculates the degree of achievement of the initial task at this point in time and presents the next action that the user should execute. Below, a process will be described in which the information processing device 10 evaluates the action executed by the user based on the user's evaluation and presents the next action that should be executed. FIG. 7 is a flowchart showing an example of the processing procedure for presenting the evaluation of an action, and FIGS. 8 to 11 are explanatory diagrams showing example screens.

[0049] After one of the actions presented to the user is selected, the control unit 11 (feedback acquisition unit 114) of the information processing device 10 displays a feedback screen as shown in FIG. 8A on the display unit 15 (S31). For example, the control unit 11 displays the feedback screen when the user instructs the display of the feedback screen. Alternatively, when user evaluation feedback is periodically collected, the control unit 11 displays the feedback screen at a predetermined timing. The screen of FIG. 8A displays a message prompting the user to enter a five-point rating for the effect of the performed action and a comment on the action. The screen of FIG. 8A has an input field R1 for the user to enter a message. The user enters evaluation information for the performed action or the action currently being performed in the input field R1 and operates the send button. The feedback acquisition unit 114 displays a message prompting the user to enter parameters for the effect obtained as a result of performing the action, depending on the content of the action selected by the user. For example, in the example of FIG. 9A, in addition to the evaluation value on a five-point rating and the evaluation comment, a message prompting the user to enter items associated with the action and obtained after performing the action (protein amount, time until sample delivery, and processing time) is displayed. The feedback acquisition unit 114 determines whether or not input of evaluation information by the user has been accepted via the input field R1 (S32), and if it determines that it has not been accepted (S32: NO), it waits until it is accepted.

[0050] When the control unit 11 (feedback acquisition unit 114, evaluation confirmation unit 115) determines that it has received user evaluation information (S32: YES), it evaluates the action performed by the user, taking into account the performance of past cases stored in the knowledge base 12a (S33). In the example of FIG. 8B, a five-point evaluation value and an evaluation comment (text data) are input. For example, the evaluation confirmation unit 115 performs natural language processing on the evaluation comment and then performs a positive / negative judgment process, and calculates a KPI score for the action performed by the user based on the positive / negative judgment result and the five-point evaluation value. The evaluation confirmation unit 115 may also update the user's environmental information collected by the initial value setting unit 111 to information after the action is performed, acquire the degree of change, and calculate a KPI score for the action performed by the user from the degree of change. The evaluation confirmation unit 115 may also acquire an expected effect of the action performed by the user from the problem decomposition unit 112, and calculate a KPI score for the action based on the performance of past cases and the expected effect.

[0051] The control unit 11 (action evaluation learning unit 116) updates the information stored in the knowledge base 12a based on the KPI scores calculated by the evaluation confirmation unit 115, and re-learns the criteria for extracting actions by referring to the knowledge base 12a (S34). The control unit 11 (evaluation confirmation unit 115) plots the calculated KPI scores on a difference product graph with the elapsed time on the horizontal axis and the KPI scores on the vertical axis (S35). The control unit 11 (evaluation confirmation unit 115) displays the difference product graph plotting the KPI scores on the display unit 15 (S36). In the example of FIG. 8B , as shown in G1, a difference product graph based on the simulation results of the action performed before the execution of the action is indicated by a dashed line, and a difference product graph plotting the KPI scores after the execution of the action is indicated by a solid line. It should be noted that, in this difference product graph G1, a difference product graph based on the simulation results of actions not selected by the user among the actions presented to the user may also be displayed using a different line type. This allows the effect (KPI score) of the action selected by the user to be compared with the effect (estimated effect) of the action not selected.

[0052] After presenting the effect (KPI score) of the action selected by the user, the control unit 11 (problem decomposition unit 112) extracts next actions the user can take to solve the original problem and achieve the goal selected by the user, or other goals for solving the original problem, generates information to be presented to the user, and displays it on the display unit 15 (S37). In step S37, the control unit 11 may perform a process similar to step S18 in FIG. 4 to extract next actions the user can take and generate action information, or may perform a process similar to step S14 in FIG. 4 to decompose the original problem and generate goal information for solving the problem. Note that here, the control unit 11 refers to the knowledge base 12a that was re-learned based on the user's evaluation in step S34, so that it can extract actions or goals based on the previous actions and their results. Furthermore, here, the control unit 11 may again engage in a dialogue with the user, create a prompt instructing the generation of an action or goal based on the content of the dialogue and the updated knowledge base 12a, and acquire the action information or goal information generated by the language model M. Furthermore, the control unit 11 may exclude from the presentable actions those actions that have been presented to the user in the past.

[0053] The control unit 11 (action presenting unit 113) determines whether the presented information is action information (S38), and if it is determined that the presented information is not action information (S38: NO), that is, if the presented information is information about a goal, it accepts the selection of one of the displayed goals via the input field R1 (S39). Then, the control unit 11 extracts actions for achieving the selected goal, generates action information (S40), and displays the generated action information on the display unit 15 (S41). Steps S39 to S41 are the same processes as steps S17 to S19 in FIG. 4.

[0054] If the control unit 11 determines that the presented information is action information (S38: YES), it skips steps S39 to S41. In the example of Fig. 8B, action information that suggests multiple companies as candidates for companies to contact is presented, as shown by M5. Here too, the problem decomposition unit 112 may determine and present a recommendation level for each of the presented companies, or may estimate and present the degree of achievement of the problem if each company is adopted.

[0055] The control unit 11 (action presentation unit 113) accepts the selection of one of the displayed actions via the input field R1 (S42). In the example of FIG. 8B, as shown by M6, a reply message notifying the user that BBB Company has been selected as the company to contact is input. The control unit 11 (action evaluation learning unit 116) performs re-learning using the user's action selection information (the event that the user selected one of the actions) as re-learning data (S43). Steps S42 to S43 are the same processes as S20 to S21 in FIG. 4. Note that in the example of FIG. 8B, the reply message shown by M6 does not notify the user of the selection of a specific action. Therefore, the control unit 11 (problem decomposition unit 112) returns to step S40, extracts specific actions that the user can take next to solve the original problem and achieve the goal selected by the user, generates action information (S40), and displays the generated action information (S41). As a result, specific actions to be taken next by the user are presented as shown by M7 in FIG. 8B. In the example of M7 in Figure 8B, the control unit 11 (problem decomposition unit 112) also displays the results of simulating the execution of each extracted action, and the simulation results here are expressed as numerical values ​​for each item (protein amount, time until delivery, processing time).

[0056] The control unit 11 (action presenting unit 113, action evaluation learning unit 116) accepts a selection of one of the displayed actions via the input field R1 (S42), and based on the action selection information by the user, re-learns the information held in the knowledge base 12a and the criteria for extracting actions by referring to the knowledge base 12a (S43). Here too, the control unit 11 performs processing such as updating the number of selections of the similar action group to which the selected action belongs in the similar action group DB of the knowledge base 12a.

[0057] As shown in FIG. 8B, a user who selects one of the actions presented by the information processing device 10 executes the selected action and then feeds back evaluation information on the executed action to the information processing device 10. In FIG. 8B, the action "send a ready-made sample" is selected, so the user sends the sample to BBB, waits for the results from BBB, and obtains the measurement results of components such as proteins extracted from the sample. Having obtained the measurement results from BBB, the user feeds back the measurement results to the information processing device 10 as the execution result (evaluation) of the action. At this time, the control unit 11 of the information processing device 10 performs the process of FIG. 7 again.

[0058] That is, the control unit 11 displays a feedback screen as shown in FIG. 9A on the display unit 15 (S31), and the user inputs the measurement results obtained from BBB as evaluation information for the action in the input field R1. The control unit 11 then receives the user's evaluation information, evaluates the action in question based on the received evaluation information and the performance of past cases held in the knowledge base 12a (S33), and calculates a KPI score for the action. The control unit 11 then plots the calculated KPI score on a difference product graph (S35), thereby presenting a difference product graph as shown in G2 in FIG. 9B. The KPI score here is obtained by plotting an arrow with a slope corresponding to the amount of change in the KPI score due to the action, with the end of the arrow in the difference product graph (shown as G1 in FIG. 8B) based on the KPI score for the most recently executed action as the base end (the start time of the action). The difference product graph shown in G2 in FIG. 9B is obtained.

[0059] The control unit 11 also executes the processes of steps S37 to S42, extracts possible actions for the user, generates and displays action information, and repeats the process of accepting a selection from the displayed actions. By repeating the above-described processes, the control unit 11 can present a difference product graph plotting KPI scores for actions executed sequentially in chronological order, as shown by G3 in FIG. 10. Such a difference product graph allows the user to grasp the effect of each action executed by the user (KPI score, degree of achievement of the initial task). The dashed line in graph G3 in FIG. 10 indicates a difference product graph estimated based on the simulation results for the action initially selected by the user. The control unit 11 may also display a difference product graph (initial presented simulation results) estimated based on the simulation results for the four actions initially presented to the user, as shown by the dashed lines in the upper left of FIG. 11A. In this case, the slope of each difference product graph allows the effect of each presented action (predicted KPI score) to be easily compared.

[0060] Furthermore, as shown by the dashed line on the right side of FIG. 11A , the control unit 11 may display a difference product graph (KPI score) estimated based on the results of a simulation of a case in which the user does not perform any of the presented actions. In this case, the effects of each performed action can be easily compared by comparing it with a case in which no action (some kind of measure) is performed. Furthermore, the control unit 11 performs a simulation for each action at the time each action is presented and estimates a KPI score based on the simulation results. Therefore, as shown by the dashed line in the lower left of FIG. 11A , the control unit 11 may display a difference product graph of the KPI score estimated based on the simulation results of each action at the time each action is presented (i.e., the time each action starts to be performed). Furthermore, as the difference product graph, as shown in FIG. 11B , the control unit 11 can generate graphs in which the vertical axis is switched between the task and multiple goals, such as a graph showing the degree of achievement (KPI score) for the initial task, a graph showing the degree of achievement (KPI score) for the first provided goal (first goal), and a graph showing the degree of achievement (KPI score) for the second goal (second goal). Therefore, the control unit 11 may display a plurality of difference product graphs side by side, or may switch between them as appropriate in response to a switching instruction from the user, for example.

[0061] By performing the above-described processing, the information processing device 10 of this embodiment infers a user's problem from information about the user's environment and presents the user with a goal for solving the inferred problem and actions for achieving the goal. The actions presented to the user are based on the ever-evolving knowledge base 12a, enabling the presentation of optimal actions. Furthermore, the information processing device 10 incorporates the concept of subtraction to present actions for solving the problem and evaluates the action selected by the user, thereby realizing the resolution of the user's problem and the achievement of the goal for solving the problem. By answering questions from the information processing device 10, the user can receive suggestions for actions to solve their problem. Furthermore, after executing a suggested action, the user's evaluation of the action is fed back to the information processing device 10, allowing the information processing device 10 to update its criteria for determining whether to present each action to the user in subsequent processing.

[0062] FIG. 12 is an explanatory diagram of a feedback process for user evaluations. If an action is performed continuously for a predetermined period of time, and the user evaluates the action once a day, evaluation information such as that shown in FIG. 12 is accumulated. In FIG. 12, the user's evaluation value, expressed on a five-point scale, is associated with the measured values ​​or the user's evaluation results (evaluation values) for the attributes associated with the action (in FIG. 12, monetary cost, time required, physical cost, communication ability, and travel distance) and this evaluation information is stored in the knowledge base 12a. For such evaluation information, for example, using a causal information analysis service, the evaluation result can be used as a target variable to derive attributes that directly affect the evaluation results. This makes it possible to change the priority of each action to be presented in subsequent processing based on the attributes. For example, if physical cost is identified as a factor that directly affects the evaluation results, feedback can be provided in subsequent processing to increase or decrease the priority of actions that require physical effort. Using the knowledge base 12a retrained by such feedback processing, the criteria for determining whether each action should be presented can be updated based on the attributes of each action.

[0063] In the above process, a configuration for using the language model M to have a dialogue with the user has been described. However, questions to the user may be registered in a table in advance. For example, for a general issue such as "I want to improve costs," items such as cost reduction, sales increase, and CO2 reduction may be registered in a table in advance. By presenting these items to the user, the user can simply select the desired item, enabling a smooth dialogue. Furthermore, by displaying a screen for inputting amounts or quantities for the items selected by the user, the amount or quantity for each item can be input via the screen, facilitating feedback processing from the user.

[0064] Furthermore, in this embodiment, the information processing device 10 is configured to extract a user's problem, and extract and present a goal for solving the extracted problem and an action for achieving the goal, but is not limited to this configuration. For example, the information processing device 10 may be configured not to perform the process of extracting a goal. In this case, the information processing device 10 may be configured to extract a user's problem, and extract and present an action for solving the extracted problem. Even with such a configuration, it is possible to execute the same process as in the above-described embodiment.

[0065] In this embodiment, the information processing device 10 is configured to extract a task from the content of an initial dialogue with the user and present various goals and actions to solve the extracted task. In addition to this configuration, after an action is performed and the user's evaluation is fed back, the task itself may be updated based on the user's evaluation, etc. This makes it possible to present actions according to the changed task, even if the task changes depending on the effect of the user's action, and to take actions to achieve a more optimal goal. Furthermore, when updating the task, the information processing device 10 may re-dialize the dialogue with the user and extract an updated task based on the user's evaluation and the content of the dialogue, or may receive permission from the user to update the task.

[0066] The features described in the above embodiments can be combined with each other. Furthermore, the independent claims and dependent claims described in the claims can be combined with each other in any and all combinations, regardless of the reference format. Furthermore, the claims use a format in which a claim references two or more other claims (multiple claim format), but this is not limited to this. A multiple claim (multi-multi claim) that references at least one other multiple claim may also be used.

[0067] The embodiments disclosed herein are to be considered in all respects as illustrative and not restrictive. The scope of the present invention is defined by the claims, not by the above meaning, and is intended to include all modifications within the meaning and scope of the claims. [Explanation of symbols]

[0068] 10. Information processing equipment 11 Control section 12 Storage section 13 Communications Department 14 Input section 15 Display M language model 12a Knowledge Base

Claims

1. Identifying a user's problem based on a dialogue with the user; Deciding on a number of actions to be taken in response to the identified issues; Estimate the evaluation of each action that has been decided upon, Output each action decided and the evaluation of each action A program that causes a computer to perform a process.

2. Identifying a plurality of goals in response to the identified challenges; Determine multiple actions to be taken for each identified goal, For each of the identified goals, estimate an evaluation of each of the determined actions; For each of the identified goals, output each determined action and an evaluation of the results of each action. The program according to claim 1 , which causes the computer to execute a process.

3. Using a language model that utilizes a knowledge database established in relation to the user's business, a plurality of actions are determined for the identified problem.

3. The program according to claim 1, which causes the computer to execute processing.

4. creating a prompt that instructs the generation of a plurality of actions for the problem based on the identified problem and extracted information extracted from the knowledge database based on the problem; The created prompt is input to the language model to obtain a plurality of actions for the task generated by the language model. The program according to claim 3, which causes the computer to execute processing.

5. receiving an evaluation by the user after performing any one of the plurality of actions; Identifying a next goal based on the received user rating and a dialogue with the user at the time the rating was received; Determine multiple actions to be taken to achieve the identified next goal. The program according to claim 2 , which causes the computer to execute processing.

6. Estimating the degree of achievement of the identified task based on the evaluation by the user; A graph showing the change over time in the achievement level for the task is generated based on each action performed in chronological order and the achievement level estimated based on the user's evaluation after each action is performed. The program according to claim 5, which causes the computer to execute processing.

7. Updating the knowledge database based on an action selected by the user from among the plurality of actions and / or an evaluation by the user of the action selected by the user. The program according to claim 3, which causes the computer to execute processing.

8. simulating each of the determined actions to estimate an evaluation score when each action is performed; Output a graph showing the evaluation score for each action 3. The program according to claim 1, which causes the computer to execute processing.

9. Estimate an evaluation score that would be obtained if none of the determined actions were taken; Output a graph showing the evaluation score if none of the above actions were taken The program according to claim 8, which causes the computer to execute processing.

10. Identifying a user's problem based on a dialogue with the user; Determine multiple actions to be taken to address the identified issues; Estimate the evaluation of each action that has been decided upon, Output each action decided and the evaluation of each action An information processing method in which processing is performed by a computer.

11. An information processing device having a control unit, The control unit Identifying a user's problem based on a dialogue with the user; Determine multiple actions to be taken to address the identified issues; Estimate the evaluation of each action that has been decided upon, Output each action decided and the evaluation of each action Information processing device.

Images