Needs-Seeds Matching Support System

Abstract

We support efficient and highly accurate needs-seed matching by utilizing generation AI. [Solution] The information processing device selects a task from the tasks defined in the task list that can be executed or assisted by the generating AI as a target task; generates a prompt for input to the generating AI based on the information provided by the user and a prompt template for the target task; provides the generated prompt to the generating AI to obtain an output; and presents the output obtained from the generating AI to the user.

Description

Technical Field

[0001] The present invention relates to a technology for assisting needs-seeds matching.

Background Art

[0002] Linking research results (technology seeds) held by research institutions and academia to business ideas (needs) of companies and local governments is called "needs-seeds matching." Appropriately linking a large number of technology seeds accumulated by previous research to specific issues in the industrial world is extremely important for the creation of innovation, the resolution of various social issues, and ultimately the enhancement of international industrial competitiveness.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] Needs-seeds matching is a complex task that requires a lot of expertise and experience, and has conventionally been carried out by experts called "coordinators." The following problems existed in the conventional coordination work.

[0005] First, there is a problem that the dependence (personal nature) on the knowledge and skills of individual coordinators is high. The success or failure of matching largely depends on the in-depth knowledge of the coordinator regarding the technical field, understanding of industry trends, communication ability, and unique know-how. For this reason, the process of the coordination work has not been standardized, and the results were likely to vary depending on the skill level of individual coordinators.

[0006] Secondly, there are challenges related to the inefficiency and human limitations of the matching process. The entire process—including collecting and analyzing vast amounts of information on needs and seeds, researching related technologies and market trends, generating application ideas, formulating hypotheses, and creating proposals—requires a great deal of time and effort. There are limitations to manual information retrieval and analysis, and it has been particularly difficult to maintain efficiency and accuracy in cross-sectoral studies spanning multiple technological fields and in matching solutions that address increasingly complex business challenges.

[0007] On the other hand, Generative Artificial Intelligence (Generative AI), which has made remarkable progress in recent years, is attracting attention for its usefulness as a tool to streamline information gathering, analysis, and exploration. Recently, attempts have begun to utilize this Generative AI for needs-seed matching.

[0008] However, because generative AI is a tool with an extremely high degree of control, there is a problem in that the goals that can be achieved and the results obtained vary depending on the user's skill. In addition, the accuracy and efficiency of matching are greatly affected by the amount or lack of information gathered and the skill of the procedure, so obtaining a stable output is not always easy.

[0009] This invention has been made in view of the above circumstances, and aims to provide a technology that supports the efficient and highly accurate matching of needs and seeds by utilizing generative AI. [Means for solving the problem]

[0010] One aspect of the present invention is a needs-seed matching support system for supporting needs-seed matching, comprising: a storage unit that pre-stores a task list in which a plurality of tasks to be performed by a user for needs-seed matching are defined; a prompt template for causing the generating AI to perform tasks that can be performed or assisted by the generating AI from among the tasks defined in the task list; and a task execution support unit that supports the performance of tasks that can be performed or assisted by the generating AI based on the task list stored in the storage unit, wherein the task execution support unit performs the following: selects tasks that can be performed or assisted by the generating AI from among the tasks defined in the task list as target tasks; generates a prompt for input to the generating AI based on information provided by the user and the prompt template for the selected target tasks; inputs the generated prompt to the generating AI and obtains an output; and presents the output obtained from the generating AI to the user.

[0011] The present invention may be understood as a needs-seed matching support system or a needs-seed matching system having at least a part of the above functions, or as a needs-seed matching support method, a needs-seed matching method, or a control method for an information processing device that includes at least a part of the above processing, or as a program for realizing such a method or a recording medium on which such a program is recorded. [Effects of the Invention]

[0012] According to the present invention, it is possible to efficiently perform highly accurate needs-seed matching by utilizing generative AI. [Brief explanation of the drawing]

[0013] [Figure 1]Figure 1 is a block diagram showing the overall configuration of the needs-seeds matching support system according to the first embodiment. [Figure 2] Figure 2 shows an example of a task list corresponding to a needs-driven approach. [Figure 3] Figure 3 is a flowchart showing an example of the operation flow of a needs-seeds matching support system. [Figure 4] Figure 4 is a flowchart showing the details of task performance support by generative AI. [Figure 5] Figures 5A and 5B show examples of screens displayed on a user terminal. [Figure 6] Figure 6 shows an example of a task progress screen displayed on the user's terminal. [Figure 7] Figure 7 shows an example of a prompt template. [Figure 8] Figure 8 shows an example of a human-based completion determination. [Figure 9] Figure 9 shows an example of automated judgment using generation AI. [Figure 10] Figure 10 is a block diagram showing the overall configuration of the needs-seeds matching support system according to the second embodiment. [Figure 11] Figures 11A and 11B show examples of keyword databases. [Modes for carrying out the invention]

[0014] The Needs-Seeds Matching Support System (hereinafter also simply referred to as the "Support System" or "this System") is a system for supporting needs-seeds matching by utilizing generative AI. This System includes a storage unit that pre-stores a task list in which multiple tasks that a user should perform for needs-seeds matching are defined, and prompt templates for causing the generative AI to perform tasks that can be executed or assisted by the generative AI from among the tasks defined in the task list, and the storage unit stores... Based on the obtained task list, there is provided a task execution support unit that supports the execution of tasks that can be executed or assisted by the generative AI. Here, the task execution support unit may select, as target tasks, tasks that can be executed or assisted by the generative AI from the tasks defined in the task list, generate a prompt for input to the generative AI based on the information provided by the user and the template of the prompt for the target tasks for the selected target tasks, input the generated prompt to the generative AI to obtain an output, and present the output obtained from the generative AI to the user.

[0015] In this system, a task list defining a plurality of tasks is prepared in advance and registered in the storage unit. As a result, the complex process of needs-seeds matching is clearly and systematically streamlined as specific "plurality of tasks" to be executed by the user. By following this task list, anyone can proceed with the matching work in a standardized procedure without depending on the experience and skills of individual coordinators, forming a basis for eliminating personal attributes and homogenizing the work quality.

[0016] When performing tasks according to the task list, regarding tasks that can be executed or assisted by the generative AI, the use of the generative AI is assisted by the task execution support unit. As a result, effective utilization of the generative AI, reduction of the user's burden, and attainment of accurate matching results can be expected. Specifically, since the task execution support unit automatically selects, as "target tasks", tasks that can utilize the generative AI from among a plurality of tasks, tasks that can utilize the generative AI can be identified without the user's own awareness or recognition. Thereby, the utilization opportunity of the generative AI is maximized. Also, since the task execution support unit automatically generates prompts, high-level prompt engineering skills are not required for the user. Therefore, the ability of the generative AI can be utilized without depending on the user's skills.

[0017] In this system, the task execution support unit may further execute a completion determination process for determining whether the selected target task has been completed based on the output obtained from the generation AI, generate a new prompt for the target task when it is determined in the completion determination process that the selected target task has not been completed, and input the generated new prompt into the generation AI to obtain a new output.

[0018] By providing the task execution support unit with the function of determining the completion of the target task, the system can evaluate and grasp the progress of task execution. As a result, for example, it becomes possible to provide a more convenient user experience, such as retry control for task execution (utilization of the generation AI for the task), progress management of needs-seeds matching, and navigation of task execution.

[0019] For example, when the target task is not completed in one response of the generation AI, with the configuration of generating a new prompt for the target task, the system can automatically generate the next step or a prompt for in-depth exploration without the user having to devise and input the next prompt again, and the interaction with the generation AI can be continued. As a result, multi-stage and continuous instructions for the generation AI become possible, and automatic guidance towards task completion can be realized even for complex tasks that are difficult to complete with a single output or tasks that require more detailed information, providing high convenience.

[0020] "Generating a new prompt for the target task" may include obtaining additional information input from the user and generating the new prompt based on the additional information input by the user.

[0021] By obtaining additional information input from the user, it becomes possible to control or modify the direction of information generation by the generation AI, dig deeper into the answer from another perspective, and draw out an output that meets the purpose towards the completion of the target task.

[0022] In this system, the task execution support unit may further perform the following actions when it determines in the completion determination process that the selected target task has been completed: determine whether there are any other incomplete tasks remaining in the task list, and if there are other incomplete tasks remaining in the task list, select those other incomplete tasks as new target tasks.

[0023] With this configuration, when one task is completed, the system automatically identifies the next task to be performed based on the task list and seamlessly transitions to its execution. This ensures that standardized procedures are executed reliably and consistently, reducing variability in coordination work.

[0024] The completion determination process may include causing the generating AI to determine, based on the output obtained from the generating AI, whether the selected target task has been completed. This enables a more sophisticated and context-aware completion determination, without human (user) intervention or relying on simple keyword matching or fixed rules, by determining whether the output of the generating AI itself matches the purpose of the task and whether sufficient information has been provided.

[0025] The completion determination process may include inputting the definition of the selected target task and the output obtained from the generating AI to the generating AI, causing the generating AI to output a score representing the degree of completion of the selected target task, and determining whether the selected target task has been completed by comparing the score output from the generating AI with a predetermined threshold.

[0026] This method allows for an accurate and objective evaluation of how well the output from the generating AI satisfies the requirements of the target task, based on a deep understanding of the task's definition and context. Furthermore, since completion can be determined automatically and mechanically, it eliminates subjectivity in completion judgments and reduces variability due to user differences.

[0027] The task execution support unit may further perform the task execution support unit by presenting the user with the progress status of the multiple tasks defined in the task list. During the needs-seeds matching process, the user can proceed with their work while understanding the current status of tasks, such as tasks currently in progress, completed tasks, and incomplete tasks.

[0028] The memory unit pre-stores a keyword database for each task, in which necessary words that should be included in the execution result of the task and unnecessary words that should not be included in the execution result of the task are defined. The task execution support unit may further perform the following actions by referring to the keyword database: determine whether the output obtained from the generation AI includes the necessary words and unnecessary words; add the necessary words to the output if it is determined that the output does not include the necessary words; and delete the unnecessary words from the output if it is determined that the output includes the unnecessary words. The reliability of the output of the generation AI can be improved by referring to the pre-prepared keyword database and modifying the output of the generation AI as needed.

[0029] <First Embodiment> The following describes a needs-seed matching support system related to one embodiment of the present invention. I will explain in detail while referring to the drawings.

[0030] (System configuration) Figure 1 is a block diagram showing the overall configuration of the needs-seed matching support system 1 according to this embodiment. This system 1 is a system for supporting needs-seed matching by utilizing the generation AI 3. Users of this system 1 can access this system 1 from a user terminal 2 via a network. As the user terminal 2, general-purpose information devices such as personal computers, smartphones, and tablet terminals can be used.

[0031] System 1 comprises, as its main components, a memory unit 10 and a task execution support unit 11. The memory unit 10 has the function of non-temporarily storing and managing programs and data used in System 1. The task execution support unit 11 has the function of supporting task execution using generated AI, and its main components include a front-end unit 12, a task management unit 13, a prompt generation unit 14, and a generated AI collaboration unit 15.

[0032] The front-end unit 12 is responsible for the interface with the user terminal 2. For example, the front-end unit 12 provides functions such as user login control, displaying the user interface (UI) on the user terminal 2, acquiring user operations and information entered by the user, and presenting the output obtained from the generating AI 3 to the user. Users can access the front-end unit 12 from a web browser or dedicated software installed on the user terminal 2 and log in to this system 1, thereby enabling them to use the needs-seeds matching support service provided by this system 1.

[0033] The task management unit 13 is responsible for managing the execution and progress of tasks, determining task completion, and controlling retries, according to a task list pre-registered in the memory unit 10. The prompt generation unit 14 is responsible for generating prompts to be input to the generation AI 3 using prompt templates pre-registered in the memory unit 10. The generation AI cooperation unit 15 is responsible for exchanging information with the generation AI 3, that is, inputting prompts to the generation AI 3 and obtaining the output (response) from the generation AI 3. Details of these functions will be described later.

[0034] System 1 is built on, for example, an information processing device such as a general-purpose computer or server, or on a cloud infrastructure, and each function is implemented in software using hardware resources such as a CPU (processor), GPU, memory, storage, and network interface. However, some or all of the functions described below may be configured by circuits (hardware) such as ASICs or FPGAs. Furthermore, multiple computers or servers may collaborate to build System 1.

[0035] Generative AI3 is a large language model (LLM) such as GPT, Gemini, Claude, or Llama. A provider is also a generator of AI3. This can be done by using the generative AI service provided by - via an API, or by using a local generative AI model built within this system 1.

[0036] (Task list) This section describes the task list used in System 1. A task list defines multiple tasks (work) that the user should perform for needs-seed matching. The process (procedure) for needs-seed matching is not uniform. For example, one approach is to set a target company or market and search for seeds that can provide solutions to the explicit and implicit needs of the target company or market (needs-based approach), or to start with seeds that the organization possesses and then search for target companies, markets, or application fields (seeds-based approach). The necessary tasks and procedures differ depending on the starting approach. Furthermore, even with the same needs-starting approach, the necessary tasks and procedures can change depending on the target client and the scale of the project. Therefore, the inventors have categorized the methods of needs-seed matching, designed appropriate tasks and procedures for each category, and formalized them as task lists. Multiple task lists for each category are pre-registered in the storage unit 10 of this system 1.

[0037] Figure 2 shows an example of a task list corresponding to a needs-driven approach. This is an example of a task list that defines a series of tasks (work) necessary from extracting and proposing useful technological seeds (results of research and development) held by the organization to the target company, until a contract is concluded. The entire coordination work is classified into eight major categories (phases): "Lead Information Processing," "Information Gathering for Proposal," "Analysis of Issues and Solutions," "Draft Proposal Creation," "Idation," "Proposal Creation," "Proposal Implementation," and "Contract Procedures," with one or more detailed tasks set within each phase. The system is designed so that appropriate needs-seed matching can be completed by the user executing multiple tasks #1 to #23 in order from top to bottom according to the task list. Here, among the 23 tasks, those with a "○" flag for "AI Support" are tasks that can be executed or assisted by generating AI (hereinafter referred to as "AI Support Tasks").

[0038] System 1 has a mechanism to guide the user so that they can perform tasks sequentially according to the task list. Furthermore, for AI-assisted tasks, it also has a function to assist the user in using the generated AI 3. A detailed explanation follows below.

[0039] (Operation Flow) An example of the needs-seed matching support process by System 1 will be explained following the flowcharts in Figures 3 and 4. Figure 3 is a flowchart showing an example of the operation flow of System 1, and Figure 4 is a flowchart detailing the task execution support by the generating AI (step S304 in Figure 3).

[0040] When a user logs into System 1 from User Terminal 2, the front-end unit 12 displays a menu screen as shown in Figure 5A. When starting a new search (needs-seeds matching), the user specifies the type of matching on this menu screen. The panel on the left side of the screen displays a list of the user's past search history (Case A, Case B, ...). By selecting a case from this history list, the user can view the results of past cases or continue working on cases where they have completed some tasks.

[0041] In step S300, if the user selects a matching type from the menu screen (i.e., a new search), in step S301, the task management unit 13 reads the task list corresponding to that type from the storage unit 10 and adds a record of the new case to the history information managed by the storage unit 10. On the other hand, if the user selects a past case from the history list in step S300, in step S301, the task management unit 13 refers to the record of the selected past case and reads the corresponding task list and task progress information. The task progress information is information that manages which tasks in the task list have been completed.

[0042] In step S302, the task management unit 13 selects an incomplete task from the task list obtained in step S301. If there are multiple incomplete tasks, it only needs to select the task that is at the very beginning of the task list. The task selected here is called the "target task".

[0043] In step S303, the task management unit 13 selects the target task in step S302. The system determines whether a task is an AI-supported task or not. Specifically, it refers to the "AI Support" flag in the task list (see Figure 2). If the flag is "○", it is determined to be an AI-supported task and the system proceeds to step S304. Otherwise, it is determined not to be an AI-supported task and the system proceeds to step S305.

[0044] In step S304, task execution support is provided by the generating AI. The details of the process in step S304 will be described in detail later using Figure 4.

[0045] In step S305, the task management unit 13 instructs the user to perform the target task via the front-end unit 12. For example, in the case of task #1 "Organize information obtained from the top executive of the proposed company" in the task list in Figure 2, the front-end unit 12 displays an instruction screen like the one in Figure 5B. The user enters the name of the proposed company, the contact information of the top executive they contacted (name, address, phone number, email address, job title, etc.), the themes the top executive was interested in (challenges they were facing, technologies they were interested in, etc.), and the circumstances and history of contacting the top executive, following the instructions on the screen, and then presses the OK button. This information is then saved to the case record in the storage unit 10. If all the information in the "required" fields has been entered, the task management unit 13 determines that the task is "completed" and updates the task progress information in the case record.

[0046] In step S306, the front-end unit 12 displays the task progress on the user terminal 2. Figure 6 is an example of a task progress screen. Completed tasks are marked with a check mark, and the next task to be performed is marked with an arrow. By checking this progress screen, the user can understand that they have completed the task "Organize information obtained from the top executives of the proposed company" and that the next task to be performed is "Research the company information of the proposed company."

[0047] In step S307, the task management unit 13 determines whether there are any other incomplete tasks remaining in the task list. If there are, it returns to step S302, selects the incomplete task as a new target task, and executes the process from step S303 onward. This process is automatically repeated until all tasks in the task list are completed.

[0048] (Step S304: Task execution support by generative AI) Figure 4 shows the detailed flow of the task execution support provided by the generation AI in step S304.

[0049] First, in step S400, the task management unit 13 reads a prompt template for the target task from the storage unit 10. Figure 7 shows an example of a prompt template stored in the storage unit 10. A predetermined prompt template is prepared for each AI support task, corresponding to its task number. In the prompt template, strings enclosed in "[" and "]" (for example, [Company Name]) are placeholders (replacement strings). The read prompt template is then passed to the prompt generation unit 14.

[0050] In step S401, the prompt generation unit 14 generates a prompt for input to the generating AI 3 based on the prompt template. At this time, the prompt generation unit 14 may complete the prompt using additional information such as information provided by the user and answers obtained from the generating AI 3 so far (hereinafter referred to as "context"). For example, the placeholder [Company Name] is replaced with the company name entered by the user on the screen in Figure 5B. Also, the placeholder [Answer for Task #8] is replaced with the prompt for Task #8. The prompt is replaced with the context obtained when the prompt was input to the generation AI 3. The context may be obtained from the response history held by the generation AI 3 itself, or from the response history recorded in the case record by the task management unit 13 each time. If the information corresponding to the placeholder included in the prompt template is unknown, the prompt generation unit 14 may request the user to input the information via the front-end unit 12. For example, if the information corresponding to the placeholder [Name of technology provider] in the prompt template for task #15 has not been provided and is unknown, an input screen saying "Please enter the name of the organization that provides the technology seed" will be displayed on the user terminal 2, and the user will be requested to enter the name of the technology provider. The prompt generated in step S401 is passed to the generation AI collaboration unit 15.

[0051] In step S402, the generation AI collaboration unit 15 inputs the generated prompt to the generation AI 3, thereby requesting the generation AI 3 to perform the target task.

[0052] In step S403, the generation AI collaboration unit 15 receives the response output from the generation AI 3. The acquired response is displayed on the user terminal 2 via the front-end unit 12 and recorded in the case record history information by the task management unit 13.

[0053] Next, in step S404, the task management unit 13 executes a completion determination process to determine whether or not the target task has been completed, based on the output (response) of the generated AI3 obtained in step S403. The completion determination may be performed by a human (user) or by an automated system.

[0054] Figure 8 shows an example of human-based completion determination. In step S403, when the front-end unit 12 displays the response of the generating AI 3 to the user terminal 2, it displays a text box along with the message, "Please enter any missing information or additional perspectives you would like to add to the above response." If the user presses the OK button without entering anything in the text box, the task management unit 13 considers that a sufficient response has been obtained and determines that the task is complete. If the task is determined to be complete (YES in step S405), the flow in Figure 4 ends and the process proceeds to step S306 in Figure 3. On the other hand, if there is missing information or additional perspectives entered in the text box, the task management unit 13 determines that the task is incomplete. If the task is determined to be incomplete (NO in step S405), the process returns to step S400, generates a new prompt for the target task (step S401), and performs the task again using the generating AI 3 (step S402). At this time, the new prompt is generated incorporating information such as "missing information" and "additional perspectives" entered by the user in the UI of Figure 8. By adding user input to prompts, it becomes possible to control or modify the direction of information generation by Generative AI3, or to delve deeper into the answer from a different perspective, thereby eliciting a better answer toward completing the target task.

[0055] Figure 9 shows an example of automated judgment using generation AI. The task management unit 13 creates a prompt for completion judgment based on the response from generation AI 3 obtained in step S403. In the example in Figure 9, the request states, "Quantitatively evaluate the degree of satisfaction of the response to the task and output a score. If there is any missing information or additional points to consider, output them in bullet points." This is followed by a definition of the target task, the response from generation AI 3 obtained in step S403, evaluation criteria, and a definition of the score output format. When such a prompt is input to generation AI 3, a score representing the degree of completion of the target task is output as a number between 0 and 100. Missing information and additional points to consider are also output. Subsequently, the task management unit 13 can determine whether the target task is complete or not by comparing the score output from generation AI 3 with a predetermined threshold. For example, if the threshold is set to 80, the example in Figure 9 (score: 65) is determined to be an incomplete task. If the task is incomplete (NO in step S405), the process returns to step S400, and a new prompt is generated for the target task. After completing the task (step S401), perform the task again using the generating AI3 (step S402). At this time, it is advisable to add the items "geopolitical risk" and "patent portfolio" pointed out by the generating AI3 to the prompt. In the case of automatic judgment, it is advisable to set an upper limit on the number of retries to prevent steps S400 to S405 from falling into an infinite loop. If the number of retries exceeds the upper limit, the answer with the highest score obtained up to that point is recorded as the final answer for the target task, and the program moves on to the next task.

[0056] <Second Embodiment> Next, a needs-seed matching support system according to the second embodiment of the present invention will be described. Hallucination is a known problem inherent to generating AI. Hallucination refers to the output of information that is not factual or information that does not exist, as if it were true. Since hallucination is unavoidable to occur with a certain probability, there is a possibility that false information may be included in the response obtained from the generating AI using the method of the first embodiment. Conversely, due to the accuracy of the generating AI, insufficient input information, or the influence of context, there may be cases in which information that should be included in the response output by the generating AI is omitted. In the second embodiment, in order to solve these problems, the reliability of the output of the generating AI is improved by referring to a pre-prepared keyword DB and correcting the output of the generating AI as necessary.

[0057] Figure 10 is a block diagram showing the overall configuration of the needs-seeds matching support system 1 according to the second embodiment. The differences from the first embodiment (Figure 1) are that the keyword DB 16 is stored in the storage unit 10 and a modification unit 17 has been added as a function of the task execution support unit 11. The other configurations are the same as those of the first embodiment, so the following explanation will focus on the differences from the first embodiment.

[0058] Keyword DB16 is a database (terminology dictionary) that defines words that should be included in the response of the generated AI3, which is the result of executing a task (hereinafter referred to as "required words"), and words that should not be included in the response of the generated AI3, which is the result of executing a task (hereinafter referred to as "unnecessary words"). It is preferable to have a separate Keyword DB16 for each task.

[0059] Figures 11A and 11B show examples of the Keyword DB16. Figure 11A is an example of the Keyword DB for Task #9 (Detailing the Issue). It defines the necessary and unnecessary words for the task of detailing the higher-level conceptual issues listed in the "Social Issues, Business Issues" column. The "Detailed Issue" column contains the words representing the detailed issue, and the "Required / Unnecessary" column indicates whether the word is required ("Required") or unnecessary ("Unnecessary"). Figure 11B is an example of the Keyword DB for Task #15 (Exploring a wide range of solution technologies in line with the direction of solutions to management issues). The technologies corresponding to the solutions listed in the "Solution Keyword" column are listed in the "Technical Approach" column. The "Required / Unnecessary" column indicates whether the word listed in "Technical Approach" is a required or unnecessary word.

[0060] In this embodiment, in steps S402 to S403 of the task execution support flow by the generation AI shown in Figure 4, the modification unit 17 performs the following processing.

[0061] In step S402, the generation AI collaboration unit 15 requests the generation AI 3 to perform the target task by inputting the generated prompt to the generation AI 3. In step S403, when the generation AI collaboration unit 15 receives the output (response) from the generation AI 3, it passes that output to the modification unit 17. The modification unit 17 refers to the keyword DB 16 for the target task stored in the memory unit 10 and determines whether the text output by the generation AI 3 contains necessary and unnecessary words. If the modification unit 17 determines that the output of the generation AI 3 does not contain the necessary words, it modifies the output of the generation AI 3 by adding those necessary words. The modification unit 17 also performs the following modification: if it determines that the output of generation AI3 contains unnecessary words, it removes those unnecessary words from the output of generation AI3.

[0062] For example, if task #9 involves detailing the social issue of "circular economy," the modification unit 17 checks whether the output of the generating AI3 includes necessary words such as "plastic recycling," "metal recycling," and "reverse supply chain" (see Figure 11A). If it finds any necessary words that are not included, the modification unit 17 makes corrections to add those necessary words.

[0063] Furthermore, if Task #15 involves searching for solution technologies that align with the solution method of "Manufacturing DX," the modification unit 17 checks whether the output of the generated AI3 includes necessary words such as "defect inspection" and unnecessary words such as "document OCR" (see Figure 11B). If the necessary word "defect inspection" is not included, the modification unit 17 makes a correction to add that necessary word. If the unnecessary word "document OCR" is included, the modification unit 17 makes a correction to delete that unnecessary word.

[0064] After these modifications are made, the output of the generated AI3 is displayed on the user terminal 2 via the front-end unit 12, and is also recorded in the case record history information by the task management unit 13.

[0065] According to the configuration of this embodiment described above, even if there is excess or deficiency of information in the output of the generated AI3, corrections are made as necessary, thereby improving the reliability of the output of the generated AI3 compared to the first embodiment.

[0066] <Advantages of this system> System 1 clearly and systematically proceduralizes the complex needs-seed matching process using a task list, guiding users to perform tasks according to the procedure outlined in the task list. This allows anyone to proceed with the matching work using standardized procedures, regardless of individual experience or skills, thereby eliminating reliance on individual expertise and ensuring uniform work quality. Furthermore, the use of generation AI by the task execution support unit 11 is expected to reduce the user's burden and lead to accurate matching results. Automatic selection of tasks that can utilize generation AI (AI-supported tasks) and automatic generation of prompts allow users to maximize the capabilities of generation AI without being aware of its use. Moreover, the capabilities of generation AI can be effectively utilized even without advanced prompt engineering skills. As a result, it is possible to efficiently perform highly accurate needs-seed matching using generation AI. [Explanation of Symbols]

[0067] 1: Needs-Seeds Matching Support System 2: User terminal 3: Generation AI 10: Storage part 11: Task Execution Support Department 12: Front end section 13: Task Management Department 14: Prompt generation unit 15: Generation AI Collaboration Department 16: Keyword Database 17: Correction section

Claims

1. A needs-seed matching support system to support needs-seed matching, A storage unit that pre-stores a task list in which multiple tasks that the user should perform for needs-seed matching are defined, and a prompt template for causing the generating AI to perform tasks that can be executed or assisted by the generating AI from among the tasks defined in the task list, The system includes a task execution support unit that assists in the execution of tasks that can be performed or assisted by the generating AI based on the task list stored in the memory unit, The aforementioned task execution support unit, From the tasks defined in the task list, select the tasks that can be executed or assisted by the generating AI as target tasks, With respect to the selected target task, a prompt for input to the generating AI is generated based on the information provided by the user and the prompt template for the target task. The generated prompt is input to the generating AI to obtain an output. The output obtained from the generated AI is presented to the user, A needs-seed matching support system characterized by performing the following:

2. The aforementioned task execution support unit, Based on the output obtained from the generated AI, a completion determination process is performed to determine whether or not the selected target task has been completed. If the completion determination process determines that the selected target task is not completed, a new prompt is generated for the target task. The generated new prompt is input to the generating AI to obtain a new output. The needs-seed matching support system according to claim 1, further characterized by performing the following.

3. Generating a new prompt for the aforementioned target task is: Obtaining additional information from the aforementioned user, To generate the new prompt based on the additional information entered by the user, The needs-seed matching support system according to claim 2, characterized by including the following.

4. The aforementioned task execution support unit, If the completion determination process determines that the selected target task has been completed, it is determined whether or not there are other incomplete tasks remaining in the task list. If there are other incomplete tasks remaining in the aforementioned task list, select those other incomplete tasks as new target tasks. The needs-seed matching support system according to claim 2, further characterized by performing the above.

5. The completion determination process described above is: Based on the output obtained from the generating AI, the generating AI is instructed to determine whether or not the selected target task has been completed. The needs-seed matching support system according to claim 2, characterized by including the following.

6. The completion determination process described above is: The definition of the selected target task and the output obtained from the generating AI are input to the generating AI, The generating AI outputs a score representing the degree of completion of the selected target task, The system determines whether the selected target task has been completed by comparing the score output from the generating AI with a predetermined threshold, The needs-seed matching support system according to claim 2, characterized by including the following.

7. The aforementioned task execution support unit, To present to the user the progress status of the multiple tasks defined in the task list, The needs-seed matching support system according to claim 1, further characterized by performing the following.

8. The storage unit pre-stores a keyword database for each task, which defines the necessary words that should be included in the execution result of the task and the unnecessary words that should not be included in the execution result of the task. The aforementioned task execution support unit, By referring to the keyword database, it is determined whether the output obtained from the generating AI contains the required words and the unnecessary words. If it is determined that the output does not contain the required word, the required word is added to the output. If it is determined that the output contains the unwanted word, the unwanted word is deleted from the output. A needs-seed matching support system according to any one of claims 1 to 7, further characterized by performing the following:

9. A control method for an information processing device that supports needs-seed matching, The information processing device has a storage unit that pre-stores a task list in which a user is required to perform multiple tasks for needs-seed matching, and a prompt template for causing the generating AI to perform tasks that can be executed or assisted by the generating AI from among the tasks defined in the task list. The control method described above is The steps include selecting a target task from among the tasks defined in the task list stored in the memory unit, which can be executed or assisted by the generating AI, The steps include generating a prompt for input to the generating AI based on the information provided by the user and a prompt template for the selected target task, The steps include inputting the generated prompt to the generating AI to obtain an output, The steps include presenting the output obtained from the generated AI to the user, A control method for an information processing device, characterized by including the following:

10. A program for causing the processor of an information processing device to execute each step of the control method described in claim 9.

Images