Information processing device, information processing method, and program

Abstract

To provide an information processing device and others, which enable automatic generation of questions for obtaining answers a user wants when a large-scale language model or the like is used.SOLUTION: An information processing device 100 is provided, comprising a reception unit 10 for receiving an input question from a user, a relevant information acquisition unit 20 for acquiring relevant information related to the user, and a request unit 30 configured to output the input question and the relevant information to a generation unit and receive an answer from the generation unit to obtain a first question.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and a program for automatically generating questions to obtain answers that users want when using a large-scale language model or the like. [Background technology]

[0002] In recent years, the use of external chatbots with artificial intelligence functions such as ChatGPT has been attracting attention. For example, in

[0041] and

[0042] of Patent Document 1, a processor 220 uses a large-scale language model 310 to generate a chatbot capable of question-and-answer (QA) on the content of a specific domain document 301, thereby generating query data 302 indicating questions expected from the given document 301, and the processor 220 uses the query data 302 generated by the language model 310 as training data to train a search model 320 for the conversational bot.

[0003] On the other hand, when using external chatbots with artificial intelligence functions such as ChatGPT, the accuracy of the answers and the appropriateness of the content vary depending on how the questions are asked, making it difficult for people who are not familiar with the system to ask appropriate questions. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Laid-Open No. 2023-076413 Summary of the Invention [Problem to be solved by the invention]

[0005] The present invention provides an information processing device, an information processing method, and a program that automatically generate questions to obtain answers that users want when using a large-scale language model or the like. [Means for solving the problem]

[0006] [Concept 1] The information processing device according to the present invention comprises: a reception unit that receives input questions from a user; a related information acquisition unit that acquires related information about the user; a request unit that outputs the input question and the related information to a generation unit and receives a response from the generation unit to obtain a first question; may also be provided.

[0007] [Concept 2] In the information processing device according to concept 1, the related information acquisition unit acquires information related to the user's affiliation, The request unit may use information relating to the affiliation of the user as the related information.

[0008] [Concept 3] In the information processing device according to concept 1 or 2, the related information acquisition unit acquires information about the user's schedule, The request unit may use information about the user's schedule as the related information.

[0009] [Concept 4] In an information processing device according to any one of concepts 1 to 3, The request unit may request the generation unit to extract first missing information that is missing in response to the first question.

[0010] [Concept 5] In the information processing device according to concept 4, The request unit may output the missing first information extracted by the generation unit, and prompt a user to input the missing first information.

[0011] [Concept 6] In the information processing device according to concept 5, The request unit may obtain the second question by outputting the first input information entered by the user corresponding to the first missing information and the first question to the generation unit and receiving an answer from the generation unit.

[0012] [Concept 7] In the information processing device according to concept 6, The request unit may prompt the user to input whether or not he or she agrees with the second question.

[0013] [Concept 8] In the information processing device according to concept 7, If the user does not agree with the second question, the request unit may request the generation unit to extract second missing information that is missing in response to the second question.

[0014] [Concept 9] In the information processing device according to concept 8, The request unit may obtain the second question by outputting the missing second missing information extracted by the generation unit, outputting second input information input by the user corresponding to the second missing information and the second question to the generation unit, and receiving an answer from the generation unit.

[0015] [Concept 10] The information processing method according to the present invention comprises: a step of receiving an input question by a reception unit; A related information acquisition unit acquires related information about the user; outputting the input question and the related information to a generation unit and receiving an answer from the generation unit to obtain a first question; may also be provided.

[0016] [Concept 11] The program according to the present invention comprises: A program installed on a computer, The computer on which the program is installed A function for accepting input questions from a user; obtaining relevant information about the user; a function of outputting the input question and the related information to a generation unit and receiving an answer from the generation unit to obtain a first question; may be realized. [Effects of the Invention]

[0017] According to the present invention, it is possible to provide an information processing device and the like that can effectively encourage behavioral changes in a plurality of users at low cost. [Brief explanation of the drawings]

[0018] [Figure 1] FIG. 1 is a block diagram for explaining an overview of an example of an information processing system according to an embodiment of the present invention. [Figure 2] FIG. 2 is a diagram showing an example of processing by the information processing system according to the embodiment of the present invention. [Figure 3] Fig. 3 is a diagram following Fig. 2, showing an example of processing by the information processing system according to the embodiment of the present invention. [A] shows [A] in Fig. 1. [F] shows [F] in Fig. 5. [Figure 4] Fig. 4 is a diagram following Fig. 3, showing an example of processing by the information processing system according to the embodiment of the present invention. [B] shows [B] in Fig. 3. [D] shows [D] in Fig. 3. [Figure 5] 5 is a diagram following FIG. 4, showing an example of processing by the information processing system according to the embodiment of the present invention. [E] shows [E] in FIG. [Figure 6A] FIG. 6A illustrates a complementary system prompt 1 in an embodiment. [Figure 6B] FIG. 6B illustrates a complementary system prompt 2 in an embodiment. [Figure 6C] FIG. 6C illustrates a complementary system prompt 3 in an embodiment. [Figure 7A] FIG. 7A illustrates an example of a system prompt 1 for indicating a problem. [Figure 7B] FIG. 7B illustrates an example of a system prompt 2 for indicating a problem. [Figure 8A] FIG. 8A illustrates a modified system prompt 1 in an embodiment. [Figure 8B] FIG. 8B illustrates a modified system prompt 2 in an embodiment. [Figure 8C] FIG. 8C illustrates a modified system prompt 3 in an embodiment. [Figure 9] FIG. 9 illustrates a user prompt (modified) in an embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0019] Embodiment "composition" Hereinafter, an embodiment of an information processing device according to the present invention will be described. In this embodiment, an information processing method using the information processing device, a program installed to generate the information processing device, and a storage medium such as a USB or DVD storing the program are also provided. Also provided are programs installed on various terminals such as personal computers, smartphones, and tablets. In this application, "or" is a concept that includes "and," and the term "A or B" means either "A," "B," or "A and B."

[0020] The information processing device 100 may be capable of communicating with a plurality of user terminals 200. The user terminal 200 may be capable of communicating with the information processing device 100 by inputting user identification information such as a login ID and a password from the user terminal 200. The user terminal 200 may have an operation unit 210 for user operations and a display unit 220 including a display screen or the like for displaying various information. In the case of a smartphone or tablet terminal, the touch panel screen serves as both the operation unit 210 and the display unit 220.

[0021] As shown in FIG. 1, the information processing device 100 of this embodiment includes a receiving unit 10 that receives an input question input by a user via an operation unit 210 of a user terminal 200, a related information acquisition unit 20 that acquires related information about the user, such as information about the user's affiliation, and a request unit 300 that requests a generation unit 300 to generate a first question using the input question and the related information. The request unit 300 may acquire the first question by outputting the input question and the related information to the generation unit 300 and receiving an answer from the generation unit 300. According to this embodiment, the generation unit 300 is requested to generate a first question using the input question and the related information, so that a first question that takes the related information into consideration can be acquired from the generation unit 300. This allows the user to obtain an appropriate answer with a simple input. Here, the generation unit 300 may utilize a machine learning model (large-scale language model) provided by another system or may be an artificial intelligence chatbot. As an example, the generation unit 300 may be an external chatbot with artificial intelligence functions, such as ChatGPT.

[0022] Examples of related information include gender, age, department, customer, etc. With regard to customers, only customers with whom there is a case such as ongoing business negotiations may be acquired by the related information acquisition unit 20. Note that, no matter what kind of question is asked, specific information may always be included as related information, and for example, information indicating to which category the user belongs (category attribute information), such as department, job title, qualifications, etc., may always be included.

[0023] The request to generate a first question to the generation unit 300 may be made by adding predetermined user information (such as gender, age, and department) to a question input by a user (input question). For example, if a user inputs, "I'm having a conference in Tokyo, so please recommend a suitable hotel for me to stay at," user attribute information such as "25 years old, male, and working in the sales department" may be added to the question input by the user, and information such as, "I'm 25 years old, male, and working in the sales department. I'm having a conference in Tokyo, so please generate a question to help me choose a suitable hotel for me to stay at," may be output to the generation unit 300. Also, if the only customer located in Tokyo is in Kanda, information such as, "I'm 25 years old, male, and working in the sales department. I'm having a conference in Kanda, Tokyo, so please generate a question to help me choose a suitable hotel for me to stay at," may be output to the generation unit 300.

[0024] An auxiliary unit 50 may be provided that generates an input question using the input question and related information. The auxiliary unit 50 may read a format (see the supplement system prompt, the indication system prompt, and the correction system prompt, which will be described later) that is pre-stored in the storage unit 60, and generate an input question by simply adding related information and information input by the user (input information) to the format.

[0025] The related information acquisition unit 20 may acquire information about the user's schedule. The entire schedule for that day may be included as related information. The request unit 30 may then request the generation of a first question using information about the user's schedule. For example, if the schedule includes information such as "Business meeting with Company A at XX at 10:00 AM," "Business meeting with Company B at △△ at 3:00 PM," and "Dinner at □□ at 7:00 PM," these information may be acquired as schedule information, and used to generate the first question. Adopting such an embodiment is advantageous in that it allows the user's schedule to be taken into consideration when requesting the generation of a question. For example, when a user inputs a plan to have a business meeting with a client this Saturday as schedule data, if the user inputs "I will have a meeting in Tokyo this weekend, so please recommend a suitable hotel for accommodation," information such as "I am 25 years old, male, and belong to the sales department. I will have a meeting with a client in Kanda, Tokyo this Saturday, so please generate questions to help me recommend a suitable hotel for accommodation" may be output to the generation unit 300. Since the name of a client is often input in a schedule, the request unit 30 may extract the client's location from the client's name based on the client database stored in the storage unit 60.

[0026] The request unit 30 may request the generation unit 300 to extract missing information (missing information) for the first question. For example, if a user inputs, "I will be holding a conference in Tokyo, so please help me find a suitable hotel," the request unit 30 may add, to the question input by the user, "I am 25 years old, male, and working in the sales department." Information such as, "I am 25 years old, male, and working in the sales department. I will be holding a conference in Tokyo, so please help me find a suitable hotel." may be output to the generation unit 300. Alternatively, a schedule may be added and the generation unit 300 may be requested to extract missing information. For example, if a user inputs, "I will be holding a conference in Tokyo, so please help me find a suitable hotel," the request unit 30 may add, to the question input by the user, "I am 25 years old, male, and working in the sales department." Information such as, "I am 25 years old, male, and working in the sales department. I will be holding a conference with a client in Kanda, Tokyo this Saturday, so please help me find a suitable hotel." may be output to the generation unit 300.

[0027] The request unit 30 may output the missing information extracted by the generation unit 300 to the user terminal, and prompt the user to input the information.

[0028] As an example, if the request unit 30 inputs information such as "I am a 25-year-old man working in the sales department. I will be having a meeting with a client in Kanda, Tokyo this Saturday, so please let me know what information I am missing to ask questions to help me choose an appropriate hotel for accommodation," the generation unit 300 will output information such as "What are the scheduled start and end times for the meeting?", "In which area of ​​Kanda will the meeting be held? Please let me know the specific address and building name if you have one," "How long will you be staying? Is it just Saturday or will it span multiple days?", "What is your planned accommodation budget?", and "Do you have any special requests for a comfortable stay? (For example, a non-smoking room, free Wi-Fi, use of a gym or pool, etc.)," ​​and the request unit 30 may output this information to the user terminal to prompt the user for input.

[0029] The request unit 30 may request the generation unit 300 to generate a second question using additional input information and the first question input by the user. The request unit 30 may read the first question stored in the storage unit 60 in association with the user identification information and the date and time. For example, if the user inputs from the operation unit 210, "The meeting is from 1:00 PM to 3:00 PM," "Kanda Station North Exit area," "XX Building," "The stay is one day," "The accommodation budget is within 10,000 yen," and "A non-smoking room and free Wi-Fi are required," the request unit 30 outputs information such as "I am a 25-year-old male working in the sales department. I will be holding a meeting with a client in Kanda, Tokyo this Saturday. Please output the questions to be input to pick an appropriate hotel for accommodation. The conditions are "The meeting is from 1:00 PM to 3:00 PM," "Kanda Station North Exit area," "XX Building," "The stay is one day," "The accommodation budget is within 10,000 yen," and "A non-smoking room and free Wi-Fi are required." to the generation unit 300.

[0030] In this manner, by outputting the missing information extracted by the generation unit 300 to the user terminal and prompting the user to input information, the information determined to be missing by the generation unit 300 is input by the user before the second question is generated, which is beneficial in that more accurate questions can be obtained from the generation unit 300.

[0031] The generator 300 outputs a second question that is deemed appropriate for the input information, and the requester 30 may prompt the user to input whether or not they agree with the second question. If the user agrees, the second question at this point is stored in the memory 60 as the final question.

[0032] On the other hand, if the user does not agree with the second question, the request unit 30 requests the generation unit 300 to extract the missing second information for the second question. The subsequent steps are the same as those described above, where the request unit 30 outputs the missing second information extracted by the generation unit 300, prompts the user for input, and requests the generation unit 300 to generate a third question using the information input by the user for the second information and the second question. The request unit 30 may then prompt the user to input whether or not they agree with the third question. If the user agrees, the third question at this point is stored in the storage unit 60 as the final question. If the user does not agree, the above steps are repeated, and a fourth question, a fifth question, ..., an nth question (where "n" is an integer) are generated.

[0033] By adopting such an embodiment, it is possible to generate an appropriate question. The generated question may be stored in the storage unit 60 in association with user identification information. Furthermore, the information stored in the storage unit 60 may be shared and used by predetermined users who belong to the same department or organization. By adopting such an embodiment, the content of an already registered question may be appropriately modified and used as a question to the generation unit 300 from the next time onwards.

[0034] A more specific example (embodiment) of this embodiment will be described with reference to FIGS.

[0035] First, the user logs in to the system of the information processing device 100 (see FIG. 2). At this time, the user may log in to the system by inputting user identification information such as a login ID and a password. When the user logs in to the system in this way, the related information acquisition unit 20 inquires about the user information and reads the user information from the user profile stored in the storage unit 60 (see "User Information Inquiry 1" in FIG. 2). The storage unit 60 also stores a complement system prompt, a suggestion system prompt, a correction system prompt, etc., which will be described later.

[0036] After logging in to the system, the user inputs a user prompt (user command) that is an input question. In this example, the user inputs the user prompt "I'm going on a business trip to Tokyo this weekend, so please recommend transportation and hotels."

[0037] The assistant unit 50 reads the complementary system prompt 1 from the storage unit 60 and inserts the user prompt into the complementary system prompt 1 to generate the complementary system prompt 2. In this embodiment, it is assumed that the complementary system prompt 1 shown in Fig. 6A is prepared. Therefore, the complementary system prompt 2 shown in Fig. 6B is generated.

[0038] The related information acquisition unit 20 inserts user attributes extracted from the user information (see "User Information Inquiry 2" in FIG. 2) into this complementary system prompt 2, and complementary system prompt 3, which is an input question, is generated (see FIG. 3). In this case, complementary system prompt 3 shown in FIG. 6C is generated. The user attributes are not limited to information related to the user, and include information such as the user's affiliation within the company and the user's schedule.

[0039] When the completion system prompt 3 is generated in this manner, the large-scale language model (LLM), which is the generation unit 300, is called (see "LLM Call 1" in FIG. 3). More specifically, the generated completion system prompt 3 is input by the request unit 30 to the large-scale language model (LLM), which is the generation unit 300, and the answer is output as a user prompt (complemented). In this example, it is assumed that the user prompt (complemented) output is "I will be on a business trip to X Company in Kanda, Tokyo on June 9th. I will depart from the Nagoya branch. Please recommend transportation and hotels." In this example, the user prompt (complemented) corresponds to the first question.

[0040] When the user prompt (completed), which is the first question, is output in this manner, the support unit 50 reads out the indication system prompt 1 from the storage unit 60 and inserts the user prompt (completed) into the indication system prompt 1 to generate the input question, indication system prompt 2. In this embodiment, it is assumed that the indication system prompt 1 shown in Fig. 7A is prepared. Therefore, the indication system prompt 2 shown in Fig. 7B is generated.

[0041] When the suggestion system prompt 2 is generated in this way, the large-scale language model (LLM), which is the generation unit 300, is called (see "LLM call 2" in FIG. 4). More specifically, the suggestion system prompt 2 is input by the request unit 30 to the large-scale language model (LLM), which is the generation unit 300, and the answer is output. In this example, "What is the budget for hotel and transportation costs?" and "What time will you be back at the office?" are output as the information (missing information) to be queried to the user.

[0042] When the information to be queried by the user is output in this manner, the information is output from output unit 80 to user terminal 200. When information related to the query (input information) is input from user terminal 200 in this manner, support unit 50 reads correction system prompt 1 from storage unit 60 and inserts the input information from the user into correction system prompt 1 to generate correction system prompt 2. In this embodiment, it is assumed that the user inputs the information "hotel stay of less than 10,000 yen," "transportation expenses of less than 17,000 yen," and "return to work at 2 p.m. on the 10th." It is also assumed that correction system prompt 1 shown in FIG. 8A has been prepared. Therefore, correction system prompt 2 shown in FIG. 8B is generated.

[0043] Then, the user prompt (completed) is inserted into this correction system prompt 2 to generate correction system prompt 3. The correction system prompt 3 shown in Figure 8C is generated. In this example, correction system prompt 3 corresponds to the second question.

[0044] Once the modified system prompt 3 is generated in this manner, the large-scale language model (LLM) that is the generator 300 is called (see "LLM Call 3" in FIG. 5). More specifically, the modified system prompt 3 is input by the request unit 30 to the large-scale language model (LLM) that is the generator 300, and the response is output as the user prompt (modified). In this example, the user prompt (modified) shown in FIG. 8D is generated.

[0045] The user prompt (modified) generated in this manner is output from the output unit 80 to the user terminal 200. If the user agrees with the prompt, the prompt (user prompt (modified)) will be used to ask a question to the question unit 300. Typically, the user prompt (modified) is input as is to the question unit 300, and an answer to the user prompt (modified) is output to the user terminal 200 via the output unit 80. As an example, when the user inputs in the user terminal 200 that he or she agrees with the user prompt (modified) (for example, by tapping an "agree" button), the user prompt (modified) may be automatically input to the question unit 300, and an answer to the user prompt (modified) may be displayed on the user terminal 200.

[0046] As another example of agreeing to a user prompt (modified), the user may be allowed to make appropriate modifications to the user prompt (modified) output to the user terminal 200, and the modified question may be automatically input into the question section 300, and the answer may be output on the user terminal 200 via the output section 80.

[0047] On the other hand, if the user does not agree with the user prompt (modified), the prompt may be used as a "user prompt (complemented)" and the previous steps (steps following "user prompt (complemented)" in FIG. 3) may be repeated. In this case, the same steps may be repeated until the user agrees with the user prompt (modified). Note that if the user does not agree with the user prompt (modified) output to the user terminal 200, the user may make appropriate modifications to the user prompt (modified) (including deleting or changing information), and the user may then use the modified user prompt as a "user prompt (complemented)."

[0048] As an example, "disagree" and "correct" may be selected, and if "disagree" is selected on the user terminal 200, the user prompt (corrected) may be used as is as the "user prompt (complemented)." On the other hand, if "correct" is selected on the user terminal 200, the user prompt corrected on the user terminal 200 may be used as the "user prompt (complemented)."

[0049] Each of the components of the present embodiment, such as the reception unit 10, the output unit 80, the related information acquisition unit 20, the request unit 30, and the auxiliary unit 50, may be realized by one or more IC chips or electronic modules, or may be realized by a circuit configuration. Furthermore, two or more "units," such as the reception unit 10, the output unit 80, the related information acquisition unit 20, the request unit 30, and the auxiliary unit 50, may be realized by one control unit.

[0050] The above-mentioned description of the embodiment and the disclosure of the drawings are merely examples for explaining the invention described in the claims, and the invention described in the claims is not limited by the above-mentioned description of the embodiment or the disclosure of the drawings. Furthermore, the description of the claims as originally filed is merely an example, and the description of the claims may be changed as appropriate based on the description in the specification, drawings, etc.

[0051] In the above-described embodiment, the description has been given mainly on the assumption that the generation unit 300 is an external device, but this is not limiting, and the information processing device 100 may have a generation unit. The generation model used by the generation unit 300 may be generated by machine learning. In this case, the question model may be generated by machine learning using input information that is an input question and related information or an input question input by a previously used user, and output information that is a first question and a second question corresponding to the input question and related information or the input question. [Explanation of symbols]

[0052] 10 Reception 20 Related Information Acquisition Department 30 Request Department 100 Information processing device 300 Generation part

Claims

1. a reception unit that receives input questions from a user; a related information acquisition unit that acquires related information about the user; a request unit that outputs the input question and the related information to a generation unit and receives a response from the generation unit to obtain a first question; An information processing device comprising:

2. the related information acquisition unit acquires information related to the user's affiliation, The information processing apparatus according to claim 1 , wherein the request unit uses information relating to the user's affiliation as the related information.

3. the related information acquisition unit acquires information about the user's schedule, The information processing apparatus according to claim 1 , wherein the request unit uses information about the user's schedule as the related information.

4. The information processing apparatus according to claim 1 , wherein the request unit requests the generation unit to extract first missing information that is missing in response to the first question.

5. The information processing device according to claim 4 , wherein the request unit outputs the first missing information extracted by the generation unit and prompts the user for input.

6. 6. The information processing device according to claim 5, wherein the request unit outputs first input information input by a user corresponding to the first missing information and the first question to the generation unit, and obtains the second question by receiving an answer from the generation unit.

7. The information processing device according to claim 6 , wherein the request unit prompts the user to input whether or not the user agrees with the second question.

8. The information processing device according to claim 7 , wherein, when the user does not agree with the second question, the request unit requests the generation unit to extract second missing information that is missing in response to the second question.

9. The information processing device described in claim 8, wherein the request unit outputs the missing second missing information extracted by the generation unit to the user, outputs second input information input by the user corresponding to the second missing information and the second question to the generation unit, and receives an answer from the generation unit to obtain the second question.

10. a step of receiving an input question by a reception unit; A related information acquisition unit acquires related information about the user; outputting the input question and the related information to a generation unit and receiving an answer from the generation unit to obtain a first question; An information processing method comprising:

11. A program installed on a computer, The computer on which the program is installed A function for accepting input questions from a user; obtaining relevant information about the user; a function of outputting the input question and the related information to a generation unit and receiving an answer from the generation unit to obtain a first question; A program that achieves this.

Images