Know-how research support device and application program

Abstract

A know-how research support device and program are provided that can obtain high-quality research results when a user conducts a research on know-how information in a specific domain. [Solution] In the know-how research support system, the know-how research support server has a question receiving unit that receives question information regarding know-how from a user's questioner terminal, a question sending unit that sends the question information to a language model 20 that has been trained with know-how documents in a specific domain, a first answer receiving unit that receives first answer information corresponding to the question information and the validity of the first answer information from the language model, a first answer sending unit that sends the first answer information and the validity of the first answer information to the questioner terminal, a question publishing unit that publishes the question information to other users in response to a publication request from the questioner terminal, a second answer receiving unit that receives second answer information to the published question information from the answerer terminal of another user, and a second answer sending unit that transmits the second answer information to the questioner terminal.

Description

[Technical field]

[0001] The present invention relates to a know-how research support device, a know-how research support program, and an application program. [Background technology]

[0002] Various know-how is accumulated in know-how sharing domain groups within companies and departments, but if know-how documents (know-how materials) are scattered, it can be time-consuming to search and investigate the information, making it difficult to find the information you are looking for, which can take up time from actual work hours.

[0003] As a related technique, for example, Patent Document 1 describes a data management device that can improve the convenience of an FAQ system. The data management device, which manages questions and answers to questions, calculates vector similarity between the questions and past logs, and extracts and searches similar past logs as answers.

[0004] For example, Patent Document 2 describes an answer search system, a data transmission / reception device, and an answer search device that allow a user to quickly obtain answer-related information for a question posed by the user without having to post the question again. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2023-120862 A [Patent Document 2] JP 2019-160334 A Summary of the Invention [Problem to be solved by the invention]

[0006] The inventions described in the above Patent Documents 1 and 2 obtain answers based on databases in which data necessary for searching answers to questions or information related to answers, such as chat logs and Q&A, has been collected and learned in advance, so if the quality or quantity of information in the database is insufficient, the questioner may not receive a sufficient answer even if he or she asks a question about the information he or she wants to know. Also, with recent generative AI, even if the quality and quantity of information in the database is sufficient, the questioner may not receive an accurate answer depending on how he or she asks the question.

[0007] The present invention has been proposed in view of the above-mentioned points, and one object of the present invention is to provide a mechanism for obtaining high-quality search results when a user conducts a search on know-how information in a specific domain. [Means for solving the problem]

[0008] In order to solve the above problems, a know-how research support device according to the present invention comprises: A know-how research support device for researching information related to know-how within a specific domain, comprising: a question receiving means for receiving question information related to the know-how from a user terminal of a user; a question sending means for sending the question information to a language model trained with know-how documents within the specific domain; a first answer receiving means for receiving first answer information corresponding to the question information from the language model; improved question receiving means for receiving improved question information obtained by improving the question information from the language model; an answer sending means for sending the first answer information and the improved question information to the user terminal; a question publishing means for publishing the improved question information to other users in response to a publication request from the user terminal; a second answer receiving means for receiving second answer information to the published improved question information from another user terminal of another user; and a second answer sending means for sending the second answer information to the user terminal.

[0009] In order to solve the above problem, the application program according to the present invention comprises: An application program for investigating information related to know-how within a specific domain, the application program causing a computer to function as a question sending means for sending question information related to the know-how to a language model that has been trained with know-how documents within the specific domain, a first answer display means for displaying first answer information from the language model to the question information and improved question information obtained by improving the question information, a disclosure request means for requesting a server to disclose the improved question information to other users, and an answer display means for displaying second answer information from other users to the improved question information disclosed on the server. Effect of the Invention

[0010] According to the embodiment of the present invention, when a user conducts a search on know-how information in a specific domain, high-quality search results can be obtained. [Brief description of the drawings]

[0011] [Figure 1] FIG. 1 is a diagram illustrating an example of a network configuration of a know-how research support system according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a diagram illustrating an example of a hardware configuration of a know-how research support server according to the present embodiment. [Diagram 3] FIG. 2 is a diagram illustrating an example of a software configuration of the know-how research support system according to the present embodiment. [Figure 4]FIG. 4 is a diagram illustrating an example of a data configuration of a DB according to the embodiment. [Diagram 5] FIG. 2 is a diagram showing a question screen example 1 of a questioner terminal according to the present embodiment. [Figure 6] FIG. 11 is a diagram showing a question screen example 2 of the questioner terminal according to the embodiment. [Figure 7] FIG. 11 is a diagram showing a question screen example 3 of the questioner terminal according to the embodiment. [Figure 8] FIG. 11 is a diagram showing a question screen example 4 of the questioner terminal according to the embodiment. [Figure 9] FIG. 2 is a diagram showing an example 1 of an answer screen of the answerer terminal according to the present embodiment. [Figure 10] FIG. 11 is a diagram showing an example 2 of an answer screen of the answerer terminal according to the present embodiment. [Figure 11] FIG. 11 is a diagram showing answer screen example 3 of the respondent terminal according to the embodiment. [Figure 12] FIG. 13 is a diagram showing an example of a question publishing screen of the questioner terminal according to the embodiment. [Figure 13] FIG. 13 is a diagram showing an example of a question screen of a questioner terminal according to the present modification. [Figure 14] FIG. 13 is a diagram showing an example of a reply screen of a reply terminal according to the present modified example. [Figure 15] FIG. 13 is a diagram showing an example of a question screen of a questioner terminal according to the present modification. [Figure 16] FIG. 13 is a diagram showing an example of a reply screen of a reply terminal according to the present modified example. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail with reference to the drawings.

[0013] <System configuration> (Network Configuration) Fig. 1 is a diagram showing an example of a network configuration of a know-how research support system according to this embodiment. The know-how research support system 100 in Fig. 1 includes a know-how research support server 10, a large-scale language model 20, a questioner terminal 30, and an answerer terminal 40, which are connected via a network 60.

[0014] The know-how research support server 10 is a server device that, when it receives a question from a questioner to obtain desired information, responds to the questioner with an answer (also called an AI answer) obtained via a large-scale language model 20 (also called a generation AI). In addition, when, for example, the questioner judges that the AI ​​answer is insufficient or the effectiveness of the AI ​​answer is low, the know-how research support server 10 publicizes (posts) a similar question to users (members) of a know-how sharing domain group within a company or department, thereby inviting answers from other users who have knowledge (hereinafter also called respondents).

[0015] Large-scale language models (LLMs)20 are language models that are trained on a knowledge base that consolidates know-how documents and know-how materials in specific domains within a company or department, in addition to general-purpose natural language processing models (LLMs) that have been pre-trained on grammar and word meanings using a large amount of text data. For example, by training the model on internal documents, it becomes possible to generate answer sentences with higher levels of expertise and know-how. For general-purpose natural language processing models, language models (generative AI) based on, for example, BERT (Bidirectional Encoder Representations from Transformers), GPT (Generative Pre-trained Transformer)-3.5, GPT-4, etc. can be used.

[0016] The questioner terminal 30 is, for example, a smartphone, a tablet terminal, or a PC (Personal Computer), and is a terminal device of a questioner who uses the know-how research support system 100. The questioner (employee, etc.) can access and log in to the know-how research support server 10 using the questioner terminal 30, and can obtain answer information on domain know-how by, for example, asking a dialogue-style question. In addition, for example, if the questioner determines that the AI ​​answer is insufficient or has low effectiveness, he or she can publicly post (post) a similar question to a predetermined group within the company or department, thereby inviting answers from respondents (other employees, etc.).

[0017] The respondent terminal 40 is, for example, a smartphone, a tablet terminal, or a PC, and is a terminal device of a respondent who uses the know-how research support system 100. Respondents (other employees, etc.) can use the respondent terminal 40 to access and log in to the know-how research support server 10, view questions that are open to replies from questioners, and write down the replies that the respondent himself / herself thinks.

[0018] The network 60 is a communication network including wired and wireless networks, and includes, for example, the Internet, a public line network, WiFi (registered trademark), Bluetooth (registered trademark), and the like.

[0019] In order to access the know-how research support server 10, a predetermined application program (a general-purpose web browser, a dedicated application, etc.) is installed in advance on the questioner terminal 30 and the respondent terminal 40.

[0020] (Hardware configuration) 2 is a diagram showing an example of a hardware configuration of the know-how research support server according to the present embodiment. The know-how research support server 10 includes a central processing unit (CPU) 11, a read only memory (ROM) 12, a random access memory (RAM) 13, a hard disk drive (HDD) 14, and a communication device 15.

[0021] The CPU 11 executes various programs and performs arithmetic processing. The ROM 12 stores programs required at startup. The RAM 13 is a work area for temporarily storing the processing by the CPU 11 and storing data. The HDD 14 stores various data and programs. The communication device 15 communicates with other devices via the network 60.

[0022] (Software configuration) FIG. 3 is a diagram showing an example of the software configuration of the know-how research support system according to this embodiment. The know-how research support server 10 has, as its main functional parts, a question receiving unit 101, a question sending unit 102, a first answer receiving unit (AI answer receiving unit) 103, a first answer sending unit 104, a question publishing unit 105, a second answer receiving unit (other user answer receiving unit) 106, a second answer sending unit 107, and a memory unit 109.

[0023] The question receiving unit 101 has a function of receiving question information about know-how in a specific domain from the questioner terminal 30 of the questioner. The question sending unit 102 has a function of sending question information to the large-scale language model 20 that has been trained on know-how documents in the specific domain.

[0024] The first answer receiving unit 103 has a function of receiving first answer information (AI answer) corresponding to question information from the large-scale language model 20. The first answer sending unit 104 has a function of sending the first answer information (AI answer) to the questioner terminal 30.

[0025] The question publishing unit 105 has a function of publishing question information to other users in response to a publishing request from the questioner terminal 30.

[0026] The second answer receiving unit (other user answer receiving unit) 106 has a function of receiving second answer information (answers from other users) to the published question information from the answerer terminals 40 of other users. The second answer sending unit 107 has a function of sending the second answer information (answers from other users) to the questioner terminal 30.

[0027] The storage unit 109 has a function of storing a user DB and a question and answer DB (FIG. 4, etc.). The user DB is a DB in which user information including questioners and answerers is registered in advance. The question and answer DB is a DB in which question information by questioners and answer information by answerers are registered.

[0028] The questioner terminal 30 has, as its main functional units, a question sending unit 301, a first answer display unit 302, a question disclosure request unit 303, and a second answer display unit 304. The question sending unit 301 has a function of sending question information related to know-how in a specific domain to the large-scale language model 20 that has been trained with know-how documents in the specific domain via the know-how research support server 10. The first answer display unit 302 has a function of displaying first answer information (AI answer) from the large-scale language model 20 to the question information via the know-how research support server 10. The question disclosure request unit 303 has a function of requesting the know-how research support server 10 to disclose the question information to other users. The second answer display unit 304 has a function of displaying second answer information from other users to the question information disclosed on the know-how research support server 10.

[0029] The respondent terminal 40 has, as its main functional sections, an open question display section 401, a second answer input section 402, and a second answer sending section 403. The public question display unit 401 has a function of displaying question information published on the know-how research support server 10. The second answer input unit 402 has a function of inputting second answer information from another user to the question information published on the know-how research support server 10. The second answer transmission unit 403 has a function of transmitting, to the know-how research support server 10, second answer information to the input question information.

[0030] Each functional unit is realized by a program executed on hardware resources such as the CPU, ROM, and RAM of the computers constituting the know-how research support server 10, the questioner terminal 30, and the respondent terminal 40. These functional units may be read as "means," "module," "unit," or "circuit." Each DB of the storage unit 109 may also be placed in an external storage device on the network 60. The computer program and application program may be stored in a computer-readable storage medium.

[0031] Furthermore, for the purpose of explanation, the functional units of the questioner terminal 30 and the answerer terminal 40 have been explained separately, but in reality, by installing one application program on the terminal, the user terminal (questioner terminal 30 and answerer terminal 40) may have all the functional units (question sending unit 301, first answer display unit 302, question disclosure request unit 303, second answer display unit 304, as well as public question display unit 401, second answer input unit 402, and second answer sending unit 403).

[0032] (Database) FIG. 4 is a diagram showing an example of the data configuration of a DB according to this embodiment.

[0033] The user DB is a DB in which user information including questioners and answerers is registered in advance. The user DB according to this embodiment has data items such as "user ID", "password", "email address", "user name", and "affiliation", for example.

[0034] "User ID" is a unique identifier assigned to each user. "User ID" and "password" are the ID and password that the user uses to access and log in to the know-how research support server 10. "Email address," "User name," and "Affiliation" are the user's email address, user name, and affiliation, respectively. Affiliation is the know-how sharing domain group to which the user belongs, such as a company or department. In this case, the "User ID" may be an employee number, etc., and the "User name" an employee name.

[0035] The question and answer DB is a DB in which question information by questioners and answer information by answerers are registered. The question and answer DB according to this embodiment has data items such as "question ID", "question date and time", "asker ID", "question title", and "question content". In addition, linked to the "question ID", it has data items such as "answer ID", "answer date and time", "answer ID", "answer content", and "status".

[0036] "Question ID" is a unique identifier assigned to each question. "Question date and time" is the date and time when the question was created (or posted) by the questioner. "Questioner ID" is the user ID of the questioner who created the question. "Question title" and "Question content" are the title and content of the question created by the questioner. "Status" indicates the public or private (closed) status of the question on the question board.

[0037] "Answer ID" is a unique identifier assigned to each answer. "Answer date and time" is the date and time when the answer was created (or posted) to the answerer. "Answer ID" is the user ID of the answerer who created the answer. "Answer content" is the content of the answer created by the answerer.

[0038] <Usage example>

[0039] (AI question screen) 5 is a diagram showing a question screen example 1 of the questioner terminal according to this embodiment. A questioner (e.g., Taro Tanaka with user ID: P001) accesses and logs into the know-how research support system 100 from the questioner terminal 30, and a "Ask AI" screen 300 shown in FIG. 5 is displayed on the display screen of the questioner terminal 30. The questioner inputs a question into a question input field 311 in the AI ​​dialogue interface, and after inputting, presses "Send" 312. At this time, the question information input into the question input field 311 is transmitted from the questioner terminal 30 to the know-how research support server 10 and further to the large-scale language model 20.

[0040] 6 is a diagram showing a question screen example 2 of the questioner terminal according to the present embodiment. When the questioner transmits question information inputted in the question input field 311, the know-how research support server 10 receives it and displays the question information in the questioner's speech bubble 313.

[0041] For example, the balloon 313 may include the following: "I want to give a presentation about product X to a new client. What kind of materials should I create?" The question information from the questioner is displayed.

[0042] 7 is a diagram showing a question screen example 3 of the questioner terminal according to this embodiment. When an AI answer to the question information of the questioner is output from the large-scale language model 20, the know-how research support server 10 receives it and displays the answer information in the speech bubble 314 of the AI.

[0043] For example, the balloon 314 may include the following: "When creating a presentation deck for Product X, it is important to consider the following steps and elements to create a presentation that will be compelling and effective in communicating information to your customers: 1. Understand your target audience · Understand their needs and expectations and tailor your materials to them. 2. Structure of presentation materials · Introduction: Provide a brief introduction to the background and importance of Product X. Problem Statement: Identify the specific challenge or problem your product solves. Benefits Description: Customer testimonials and success stories: By including existing customer testimonials and success stories, you can highlight the reliability and track record of your product. Past success stories include Company A, Company B, and Company C. Company A is... The AI's response information is displayed.

[0044] In this embodiment, for example, the validity 315 of the AI ​​answer is displayed below the AI's speech bubble 314. The validity 315 is a value indicating the validity of the AI ​​answer, such as "The validity of this answer is 78 points." The validity is the validity of the AI ​​answer calculated by the generating AI (large-scale language model 20) itself. When the know-how research support server 10 transmits question information to the large-scale language model 20, an instruction to also answer the validity of the answer is added to the prompt. More specifically, when the generating AI is instructed to "Answer this question. Also, please provide the validity of your answer," the generating AI can calculate the validity of the AI ​​answer together with the AI ​​answer. Note that this validity calculation method depends on the specifications of the generating AI, but generally, it calculates an estimated number such as "If you answer like this in this context, the validity will be about this much," or "The match (similarity) between the question and the knowledge base is about this much, so the validity will be about this much."

[0045] After the AI's speech bubble 314 is displayed, if the questioner judges that the AI's response has sufficiently answered the question or that the validity 315 of the AI's response is high, the questioner's question can be considered resolved and the question can be closed by pressing "end question" 316.

[0046] On the other hand, if the asker judges that the AI ​​answer is insufficient or that the validity 315 of the AI ​​answer is low, the asker considers that the question of the asker is not resolved and presses "Post Question" 317. This enables the asker to publish a similar question on the question bulletin board to know-how sharing domain groups within the company or department, and to solicit better answers from other users (answerers) who have knowledge.

[0047] The AI ​​answer validity 315 is a guideline. Even if the AI ​​answer validity 315 is low, if the questioner judges the AI ​​answer to be sufficient, the question may be closed. Conversely, even if the AI ​​answer validity 315 is high, if the questioner judges the AI ​​answer to be insufficient, the question may be made public.

[0048] When the questioner presses "Post a question" 317, the entered question information is sent from the questioner terminal 30 to the know-how research support server 10, and a new record including "Question ID," "Question date and time," "Questioner ID," "Question title," "Question content," and "Status" is stored in the question and answer DB (Figure 4).

[0049] FIG. 8 is a diagram showing a question screen example 4 of the questioner terminal according to the present embodiment. When the questioner presses "Post question" 317, "Question title" 318 and "Question content" 319 to be published on the question bulletin board are displayed for confirmation. "Question content" 319 is a question by the questioner, that is, question information entered in question input field 311 and displayed in balloon 313. "Question title" 318 is the question title entered in question input field 311, and for example, a predetermined number of characters excerpted from the question information in "Question content" 319 are initially entered. Note that "Question content" 319 may be generated by a generation AI or a title or summary based on the question information, or may be entered and edited by the questioner. When the questioner publishes (posts) "Question title" 318 and "Question content" 319, the questioner can transmit a request for publishing the question information on the question bulletin board to the know-how research support server 10 by pressing "Publish" 320, and publish the question in the know-how sharing domain group. The asker may be allowed to arbitrarily specify the target of the question disclosure, such as "the company," "the department," or a specific user.

[0050] Fig. 9 is a diagram showing an example 1 of a reply screen of a reply terminal according to this embodiment. A replyer (for example, Suzuki Ichiro, user ID: P002) accesses and logs into the know-how research support system 100 from a reply terminal 40, and a "question bulletin board" screen 400 shown in Fig. 9 is displayed on the display screen of the reply terminal 40. In the "question bulletin board", a list 411 of questions posted by questioners is displayed together with "selection", "question ID", "question date and time", "question title" and "number of answers" for each question. A replyer checks "selection" for a desired question and presses "question details" 412 to move to a details screen of that question.

[0051] Fig. 10 is a diagram showing an example 2 of an answer screen of the answerer terminal according to this embodiment. The question details screen displays the "Question title" 413 and "Question content" 414 for the question selected by the answerer. Needless to say, "Question title" 413 and "Question content" 414 correspond to "Question title" 318 and "Question content" 319 made public by the asker (Fig. 8). When the answerer refers to "Question title" 413 and "Question content" 414 and answers the question, the answerer presses "Answer" 415 to transition to the answer screen for the question.

[0052] When the questioner makes public (posts) the "question title" 318 and "question content" 319, the know-how research support server 10 also stores the AI ​​answer (e.g., answer information in the AI's speech bubble 314) in the question and answer DB. Then, on the answerer screen of the answerer terminal (such as the question details screen shown in FIG. 10), the saved AI answer may be displayed in addition to the "question title" 413 and "question content" 414 for the question selected by the answerer. This can be used as a reference when the answerer answers the "question content" 414.

[0053] Furthermore, if the AI ​​answer has been saved, the answerer may input answer information for the "question content" 414 by editing the AI ​​answer as the base. In other words, the answerer can shorten the answering time by maintaining the sufficient parts of the AI ​​answer and brushing up the AI ​​answer only in the insufficient parts.

[0054] 11 is a diagram showing an example 3 of an answer screen of the answerer terminal according to this embodiment. The answer screen for a question displays a "question" 414 by the asker and a "your answer" 416. The answerer inputs (writes) answer information to the "question" 414 in the "your answer" 416 field, and after inputting, presses "send" 312.

[0055] For example, in "Your Answer" 416, "When creating a pitch deck for Product X, it's important to consider the following steps and elements to create a deck that will be compelling and effective in communicating information to your customers: 1. Understand your target audience Understand your customer needs. Find out what products, if any, they currently use. 2. Structure of presentation materials · Introduction: Provide a brief introduction to the background and importance of Product X. · Problem statement: We clearly state the specific problem or issue that the product can solve. If there is a product currently being used by the customer or a rival product, we also compare it with that product. Benefits Description: Customer testimonials and success stories: By including existing customer testimonials and success stories, you can highlight the reliability and track record of your product. A notable example of a past success story is that of B Co., Ltd. B Co., Ltd. is... The answer information is entered based on the answer:

[0056] The respondent who answered the question 414 (for example, User ID: P002 Suzuki Ichiro) is a user with knowledge in the know-how sharing domain group. For example, the respondent's answer goes one step further than the AI ​​answer of "If there are products currently being used by customers or rival products, we will also compare them with those products." In addition, the respondent's answer of "Past success cases are notable for the case of B Co., Ltd." is more in line with the actual experience of the domain group than the AI ​​answer of "Past success cases include cases of A Co., Ltd., B Co., Ltd., C Co., etc." In this regard, the AI ​​answer was indeed understood from the internal documents that there were indeed cases of A Co., Ltd., B Co., Ltd., C Co., etc. as past success cases in the pre-trained knowledge base (domain know-how documents), but the respondent's answer was the result of judging that the case that should be taken up as a past success case and paid attention to is particularly B Co., Ltd.

[0057] When the respondent presses "Send" 312, the entered answer information is sent from the respondent terminal 40 to the know-how research support server 10, and a new record including the "Answer ID," "Answer date and time," "Answer ID," and "Answer content" linked to the "Question ID" of the question is stored in the question and answer DB (Figure 4).

[0058] Fig. 12 is a diagram showing an example of a question public screen of a questioner terminal according to this embodiment. The questioner (e.g., user ID: P001, Taro Tanaka) can refer to question information 321 that he / she has publicly posted on the bulletin board on "Your Public Questions" screen 300 shown in Fig. 12. Furthermore, when an answer is given to the question from an answerer, the answer information can be referred to. Specifically, on screen 300 shown in Fig. 12, one answer has been given from an answerer (e.g., user ID: P002, Ichiro Suzuki) to the question of the questioner (e.g., user ID: P001, Taro Tanaka), and the questioner can refer to answer content 322.

[0059] If the questioner judges that a sufficient answer has been obtained from the answerer, he / she presses "Yes" 323 to close the question. On the other hand, if he / she judges that a sufficient answer has not been obtained, he / she presses "No" 324 to keep the question open on the question bulletin board. When the questioner presses "Yes" 323, the know-how research support server 10 updates the "status" of the question information in the question and answer DB to private (FIG. 4). Note that if the questioner judges that a sufficient answer has been obtained from the answerer, the know-how research support server 10 can re-learn (feed back) the question information and the answer information into the large-scale language model 20, thereby improving the accuracy of the AI ​​answer.

[0060] As described above, a user (questioner) can quickly obtain information by searching for know-how (for example, searching for internal documents) through a dialogue (Q&A) with the generating AI, but the AI's response may sometimes be inappropriate or insufficient. According to this embodiment, when an inappropriate or insufficient response is obtained from the generating AI, the user (questioner) can publicly post the same question to other users (answerers) to solicit their knowledge (by asking the knowledgeable person again), and can obtain an appropriate and sufficient response with the cooperation of other users (answerers) who have knowledge. In other words, according to the know-how research support system 100 according to this embodiment, the user can quickly obtain higher quality know-how information from both the aspects of know-how search by the generating AI and information provision by knowledgeable people.

[0061] <Variation 1> Fig. 13 is a diagram showing an example of a question screen of a questioner terminal according to this modification. This is a modification of the question screen of the questioner terminal shown in Fig. 8. As described above, if the questioner judges that the AI ​​answer is insufficient or that the validity 315 of the AI ​​answer is low, the questioner can publish a similar question in a know-how sharing domain group within the company or department, and solicit better answers from other users with knowledge.

[0062] Here, the example question screen of the questioner terminal according to this modification has, in addition to "question title" 318 and "question content" 319 made public to the questioner, "question content improved by AI" 331. Question content 331 is question information that has been improved and created by the generation AI based on "question title" 318 and "question content" 319, with the aim of making it easier for the answerer to respond more appropriately to the published question and obtain a better answer (a high-quality answer).

[0063] For example, "Question content" 319 is "I want to give a presentation about product X to a new customer who I have made an appointment with. What kind of materials should I create?", whereas "Question content improved by AI" 331 displays the question information "I want to give a presentation about product X to a new customer who I have made an appointment with. The audience is the management. What kind of materials should I create?" Comparing the two, the question information improved and created by the generation AI based on "Question title" 318 and "Question content" 319 has "The audience is the management." added. In this case, for example, by clearly indicating the audience (target) to whom the presentation will be given in the question, it becomes clear for whom the respondent should create the presentation materials, and the questioner can obtain a better answer (higher quality answer) from the respondent.

[0064] When the questioner wants to publish (post) the “Question title” 318 and the “AI-improved question content” 331, the questioner can publish the AI-improved question in the know-how sharing domain group by pressing “Publish” 332.

[0065] In recent years, the generation AI may not be able to obtain an accurate answer depending on how the questioner asks the question ("question content" 319), so the generation AI improves the question to make up for any deficiencies or inadequacies in the "question content" 319, and the questioner can publish the "question title" 318 and the "AI-improved question content" 331 on the question bulletin board instead of the "question content" 319. This allows the answerer to know the background of the questioner's question and provide a more appropriate answer to the posted question.

[0066] On the other hand, when the questioner wants to make public (post) the “question title” 318 and the original “question content” 319, he / she can press “Make public the question you created” 333 to make the question public / post in the know-how sharing domain group.

[0067] The know-how research support server 10 adds a command to the generation AI (large-scale language model 20) to create a better improvement proposal for the question information in the prompt, along with the question information. More specifically, when the generation AI is commanded to "improve the question to obtain a better answer," the generation AI responds with "AI-improved question content" 331 to the know-how research support server 10. Also, in the field of generation AI, there is something called a goal-seeking prompt, and a method of having the generation AI brush up on a prompt (question) is also known.

[0068] 14 is a diagram showing an example of an answer screen of an answerer terminal according to this modification. The question details screen displays the "Question title" 413 and the "Question content" improved by AI 421 for the question selected by the answerer. "Question title" 413 and "Question content" 414 correspond to "Question title" 318 and "Question content improved by AI" 331 in FIG. 13.

[0069] <Variation 2> FIG. 15 is a diagram showing an example of a question screen of a questioner terminal according to this modification. This is a modification of the question screen of the questioner terminal shown in FIG. 8. In the example of the question screen of the questioner terminal according to this modification, in addition to a "question title" 318 and a "question content" 319 that are made public to the questioner, there is also hearing information 341 for the questioner and an answer input field 342 for the questioner. The hearing information 341 is a hearing matter that is created by the generation AI based on the "question title" 318 and the "question content" 319 for the purpose of making it easier for the answerer to answer the published question more appropriately and to obtain a better answer (a high-quality answer), and is used to supplement the information that is lacking in the "question content" 319. The questioner inputs an answer to the hearing information 341 in the answer input field 342 and presses "Publish" 343, whereby the hearing information 341 by the questioner and its answer (question supplement information) are made public in the know-how sharing domain group together with the "question title" 318 and the "question content" 319.

[0070] In recent generation AIs, depending on how the questioner asks the question ("question content" 319), there are cases where an accurate answer cannot be obtained, so the generation AI creates hearing information 341 to supplement the deficiencies in "question content" 319 and has the questioner supplement the information that is missing in "question content" 319. The hearing information 341 and the supplemental answer are then published on the question bulletin board as question supplemental information together with "question title" 318 and "question content" 319, thereby supplementing the information that is missing in "question content" 319 by the questioner. This allows the answerer to learn the background of the question asked by the questioner and provide a more appropriate answer to the posted question.

[0071] On the other hand, if the questioner does not want to publish (post) the question supplementary information based on the hearing information 341 and the supplementary answers, he / she can publish only the question that he / she created in the know-how sharing domain group by pressing “Publish only the question” 344.

[0072] The know-how research support server 10 adds to the generation AI (large-scale language model 20) a command to create a hearing sentence to allow the questioner to fill in the missing information in the prompt along with the question information. More specifically, when the generation AI is commanded to "If there is any information missing in the question information, please create a hearing sentence to allow the questioner to fill in the missing information," the generation AI responds with the hearing information 341 to the know-how research support server 10.

[0073] FIG. 16 is a diagram showing an example of an answer screen of an answerer terminal according to this modification. The question details screen displays the "question title" 413, the "question content" 414, and the "question supplementary information" 422 for the question selected by the answerer. The "question supplementary information" 422 corresponds to the hearing information 341 and the answer entered in the answer input field 342 in FIG. 15. For example, the "question supplementary information" 422 shown in FIG. 16 displays the following question supplementary information: "1. Your business type or field is IT or system development." and "2. There is no need to provide related market research data or sources." By supplementing the missing information in the question information with the question supplementary information in this way, the answerer can understand the question background, the asker's intentions and requests, etc., and the asker can obtain a better answer (high-quality answer) from the answerer.

[0074] <Summary> As described above, the user (questioner) can quickly obtain information by searching for know-how through a dialogue with the generating AI, but the AI's response may sometimes be inappropriate or insufficient. According to this embodiment, when the user (questioner) receives only an inappropriate or insufficient response from the generating AI, the user (questioner) can publicly post the same question to other users (answerers) to solicit their knowledge (by asking the knowledgeable person again), and can obtain an appropriate and sufficient response with the cooperation of other users (answerers) who have knowledge. In other words, according to the know-how research support system 100 according to this embodiment, the user can quickly obtain higher quality information from both the know-how search by the generating AI and the information provided by the knowledgeable person.

[0075] Although the present invention has been described by showing specific examples according to the preferred embodiment of the present invention, it is apparent that various modifications and changes can be made to these examples without departing from the broad spirit and scope of the present invention as defined in the claims. In other words, the details of the examples and the accompanying drawings should not be construed as limiting the present invention. [Explanation of symbols]

[0076] 10 Know-how research support server 20 Large-scale language models 30 Questioner terminal 40 Respondent terminal 60 Network 100 Know-how research support system 101 Question Reception Department 102 Question sending section 103 First Reply Receiving Department 104 First Reply Transmission Department 105 Questions and Answers Section 106 Second Reply Receiving Department 107 Second Reply Transmission Department 109 Storage section 301 Question Sending Department 302 1st answer display section 303 Questions Publication Request Department 304 2nd answer display section 401 Public Questions Display Section 402 Second Answer Input Section 403 Second reply sending section

Claims

1. A know-how research support device for researching information related to know-how within a specific domain, comprising: A question receiving means for receiving question information regarding the know-how from a user terminal of a user; a question sending means for sending the question information to a language model trained on know-how documents in the specific domain; a first answer receiving means for receiving, from the language model, first answer information corresponding to the question information; an improved question receiving means for receiving improved question information obtained by improving the question information from the language model; an answer transmitting means for transmitting the first answer information and the improved question information to the user terminal; a question publishing means for publishing the improved question information to other users in response to a publishing request from the user terminal; a second answer receiving means for receiving second answer information to the published refined question information from another user terminal of another user; a second response sending means for sending the second response information to the user terminal; A know-how research support device comprising:

2. A know-how research support device for researching information related to know-how within a specific domain, comprising: A question receiving means for receiving question information regarding the know-how from a user terminal of a user; a question sending means for sending the question information to a language model trained on know-how documents in the specific domain; a first answer receiving means for receiving, from the language model, first answer information corresponding to the question information; A response sending means for sending the first response information to the user terminal; a hearing information receiving means for receiving hearing information from the language model to allow a user to supplement information that is lacking in the question information; a question publishing means for publishing the question information and supplemental question information for the hearing information to other users in response to a publishing request from the user terminal; a second answer receiving means for receiving second answer information to the published question information from another user terminal of another user; a second response sending means for sending the second response information to the user terminal; A know-how research support device comprising:

3. An application program for investigating information related to know-how within a particular domain, comprising: Computer, a question sending means for sending question information about the know-how to a language model trained on know-how documents in the specific domain; a first answer display means for displaying first answer information from the language model to the question information and improved question information obtained by improving the question information; a disclosure request means for requesting the server to disclose the improved question information to other users; an answer display means for displaying second answer information from other users to the improved question information published on the server; Application programs that allow the device to function as a network device.

4. An application program for investigating information related to know-how within a particular domain, comprising: Computer, a question sending means for sending question information about the know-how to a language model trained on know-how documents in the specific domain; a first answer display means for displaying first answer information from the language model to the question information and hearing information for allowing a user to supplement information that is lacking in the question information; a disclosure request means for requesting the server to disclose the question information and supplemental question information for the hearing information to other users; a response display means for displaying second response information from other users to the question information published on the server; Application programs that allow the device to function as a network device.

Images