Question determination device and question determination method

Abstract

This question determination device includes: a category determination unit that, on the basis of at least one answer candidate related to an inquiry from a user, determines at least one category to which said at least one answer candidate belongs; a related category determination unit that determines, from among said at least one category, at least one related category which is related to the inquiry; and a question determination unit that determines a question pertaining to the purpose of the inquiry, by using said at least one related category.

Description

Question Determination Device and Question Determination Method

[0001] The present disclosure relates to a technique for determining questions.

[0002] Patent Document 1 discloses an interactive management system that automatically answers inquiries from users. This interactive management system stores answer candidates having a hierarchical structure. The answer candidates are attached with classification symbols indicating hierarchies such as "Chapter 1, Section 2". When an inquiry from a user is ambiguous, this interactive management system gives the user a question including a classification symbol.

[0003] Japanese Patent Application Laid-Open No. 2019-144755

[0004] The interactive management system described in Patent Document 1 gives the user a question including a classification symbol such as "Chapter 1, Section 2". The classification symbol such as "Chapter 1, Section 2" included in the question is likely to be unnecessary information for the user who made the inquiry. Therefore, the interactive management system described in Patent Document 1 is likely to make the question redundant.

[0005] An object of the present disclosure is to provide a technique capable of reducing the possibility of a question becoming redundant.

[0006] A question determination device according to an aspect of the present disclosure includes a category determination unit that determines at least one category to which the at least one answer candidate belongs based on at least one answer candidate related to an inquiry from a user, a related category determination unit that determines at least one related category related to the inquiry from among the at least one category, and a question determination unit that determines a question related to the purpose of the inquiry by using the at least one related category.

[0007] A question determination method according to another aspect of the present disclosure includes determining at least one category to which the at least one answer candidate belongs based on at least one answer candidate related to an inquiry from a user, determining at least one related category related to the inquiry from among the at least one category, and determining a question related to the purpose of the inquiry by using the at least one related category.

[0008] This disclosure provides a technology that can reduce the likelihood of questions becoming redundant.

[0009] This figure shows an example of the response system RS. This figure is intended to illustrate the general operation of the response system RS. This figure shows an example of the terminal 10. This figure shows an example of the server 20. This figure shows an example of the answer candidate group C1. This figure shows an example of the answer candidate group C2. This figure shows an example of the functional configuration of the processing unit 23. This figure shows an example of prompt F1 indicating an instruction to generate at least one objective label A5 based on query A1. This figure shows an example of prompt F2 indicating an instruction to generate at least one objective label A7 based on update query A4. This figure shows an example of at least one objective label A5. This figure shows an example of at least one objective label A7. This figure shows an example of question A21 inquiring about the purpose of the query. This figure shows an example of question A22 confirming the purpose of the query. This figure shows an example of prompt F3 indicating an instruction to generate a sentence based on query A1, question A2, and answer A3. This figure shows an example of the question management unit 233. This figure shows an example of prompt F4 indicating an instruction to generate a summary of query A1. This figure shows an example of summary A8. This figure shows an example of prompt F5 indicating an instruction to generate at least one category A10 to which at least one manual A9 belongs. This figure shows an example of at least one category A10. This figure shows an example of prompt F6 indicating that relevance information A12 is generated based on at least one category A10 and a summary A8 of query A1. This figure shows an example of relevance information A12. This figure shows an example of prompt F7 indicating that question A21 is generated based on multiple relevance categories. This figure shows an example of template T1. This figure shows another example of question A21. This figure shows an example of prompt F8 indicating that question A22 is generated based on one relevance category. This figure shows an example of template T2. This figure shows another example of question A22. This figure illustrates the operation of server 20. This figure illustrates the operation of server 20.

[0010] 1: Embodiment 1-1: Response System RS Figure 1 shows an example of the response system RS. The response system RS automatically responds to inquiry A1 from user U1. Inquiry A1 is, for example, an inquiry about service E1. Service E1 is, for example, a communication service. Service E1 is not limited to a communication service. For example, service E1 may be a service about points that can be used for payment. Inquiry A1 is not limited to an inquiry about service E1. For example, inquiry A1 may be an inquiry about a product. The response system RS includes a terminal 10, a server 20, and a generation AI (Artificial Intelligence) system 30.

[0011] Terminal 10 is, for example, a smartphone. Terminal 10 is not limited to a smartphone. For example, terminal 10 may be a tablet or a personal computer. Terminal 10 is used by user U1.

[0012] Server 20 is composed of multiple computers. Server 20 may also be composed of a single computer. Server 20 and terminal 10 can communicate with each other via a communication network NW. Server 20 and terminal 10 may also communicate with each other without using a communication network NW.

[0013] The generation AI system 30 is composed of multiple computers. The generation AI system 30 may be composed of a single computer. The generation AI system 30 functions as a generation AI 30a. The generation AI 30a includes, for example, a Large Language Model (LLM). The generation AI 30a generates information such as text in response to a prompt. The generation AI 30a can communicate with the server 20 via a communication network NW. The generation AI 30a may also communicate with the server 20 without using a communication network NW. The generation AI 30a may also communicate with the terminal 10. The generation AI 30a is used by the server 20. The generation AI 30a may be included in the server 20.

[0014] Figure 2 is a diagram illustrating the overview of the operation of the response system RS. Terminal 10 receives inquiry A1 from user U1. Inquiry A1 is, for example, "Tell me about service E1". Terminal 10 provides inquiry A1 to server 20. For example, terminal 10 provides server 20 with a message indicating inquiry A1.

[0015] Server 20 receives inquiry A1 from terminal 10. Based on inquiry A1, Server 20 determines question A2 regarding the purpose of inquiry A1. For example, if the purpose of inquiry A1 is ambiguous, Server 20 determines question A2 by using generation AI 30a. For example, Server 20 determines question A2 as question A2, based on a prompt from inquiry A1, generated by generation AI 30a. Question A2 is, for example, "Do you want to know about a new contract for service E1? Or do you want to know an overview of service E1?" Details of the method for determining question A2 will be described later. Server 20 provides question A2 to terminal 10. For example, Server 20 provides terminal 10 with a message indicating question A2.

[0016] Terminal 10 receives question A2. Based on receiving question A2, terminal 10 displays question A2. Terminal 10 receives a response A3 to question A2 from user U1. Response A3 is, for example, "I would like to know about a new contract for service E1." Terminal 10 provides response A3 to server 20. For example, terminal 10 provides server 20 with a message indicating response A3.

[0017] Server 20 receives response A3 from terminal 10. Based on inquiry A1, question A2, and response A3, Server 20 determines the purpose of inquiry A1.

[0018] For example, server 20 first determines update query A4 based on query A1, question A2, and response A3. Update query A4 is, for example, a query obtained by updating query A1 based on query A1, question A2, and response A3. Update query A4 is, for example, a query such as "Please tell me about a new contract for service E1." For example, server 20 determines update query A4 by using generation AI 30a. To give one example, server 20 determines the sentence generated by generation AI 30a in response to prompts based on query A1, question A2, and response A3 as update query A4.

[0019] Next, the server 20 determines the purpose of query A1 based on the update query A4. For example, the server 20 determines the purpose of query A1 by using the generation AI 30a. To give one example, the server 20 first causes the generation AI 30a to generate candidate purposes for query A1 in response to a prompt based on the update query A4. Then, the server 20 determines the purpose of query A1 based on the candidate purposes for query A1.

[0020] In the response system RS, new questions and new answers are added until the probability that the determination result of the purpose of inquiry A1 is correct exceeds the threshold TH1. When new questions and new answers are added, the server 20 determines the purpose of inquiry A1 based on inquiry A1, question A2, answer A3, the new questions and new answers.

[0021] As shown in Figure 1, the server 20 stores multiple manuals C. These manuals C are used for various answers. The manuals C are an example of multiple answer candidates. The manuals C are categorized according to the purpose of the query.

[0022] Server 20 selects a manual from among several manuals C that corresponds to the purpose of inquiry A1. Server 20 responds to inquiry A1 according to the manual corresponding to the purpose of inquiry A1. Therefore, in order to respond appropriately to inquiry A1, it is important for Server 20 to correctly determine the purpose of inquiry A1. Server 20 determines the purpose of inquiry A1 based on inquiry A1, question A2, and response A3. Therefore, Server 20 can correctly determine the purpose of inquiry A1 compared to a configuration that determines the purpose of inquiry A1 based only on inquiry A1.

[0023] Server 20 provides the terminal 10 with an answer to inquiry A1, according to the manual for the purpose of inquiry A1. For example, Server 20 provides the terminal 10 with a message indicating the answer to inquiry A1.

[0024] Terminal 10 receives the answer to inquiry A1. Based on the receipt of the answer to inquiry A1, terminal 10 displays the answer to inquiry A1.

[0025] Note that Inquiry A1, Question A2, Answer A3, and Update Inquiry A4 are not limited to the example shown in Figure 2 and can be modified as appropriate.

[0026] 1-2: Terminal 10 Figure 3 shows an example of terminal 10. Terminal 10 includes an input device 11, a display device 12, a communication device 13, a storage device 14, a processing device 15, and a bus 16.

[0027] Bus 16 is a wiring configuration for communicating information. Bus 16 connects the input device 11, the display device 12, the communication device 13, the storage device 14, and the processing device 15 to each other. Bus 16 may consist of a single bus, or it may consist of various buses provided between each device.

[0028] The input device 11 is the user interface of the terminal 10. The input device 11 includes a touch panel. In addition to the touch panel, the input device 11 may include a plurality of operation keys. The input device 11 may include a plurality of operation keys without including a touch panel. The input device 11 may include a voice input device. The input device 11 receives various inputs from the user U1. For example, the input device 11 receives inquiry A1 and response A3 separately from the user U1.

[0029] The display device 12 displays various information. For example, the display device 12 displays inquiry A1, question A2, and answer A3 individually. The display device 12 may also display inquiry A1, question A2, and answer A3 together.

[0030] The communication device 13 can communicate with the server 20 via the communication network NW. The communication device 13 may also communicate with the server 20 without using the communication network NW. The communication device 13 sends inquiry A1 to the server 20. The communication device 13 receives question A2 from the server 20. The communication device 13 sends reply A3 to the server 20.

[0031] The storage device 14 is a recording medium that can be read by the processing device 15. The storage device 14 includes at least one memory. The storage device 14 includes, for example, non-volatile memory and volatile memory. Non-volatile memory is, for example, ROM (Read Only Memory), EPROM (Erasable Programmable Read Only Memory), and EEPROM (Electrically Erasable Programmable Read Only Memory). Volatile memory is, for example, RAM (Random Access Memory) and VRAM (Video Random Access Memory). The storage device 14 stores program PG1. Program PG1 includes at least one instruction.

[0032] The processing unit 15 includes at least one CPU (Central Processing Unit). The at least one CPU is an example of at least one processor. The at least one processor is an example of at least one computer. The processing unit 15 reads the program PG1 from the storage device 14. The processing unit 15 functions as an operation control unit 151 by executing the program PG1.

[0033] The operation control unit 151 controls the operation of the terminal 10. For example, the operation control unit 151 causes the communication device 13 to send inquiry A1 to the server 20. The operation control unit 151 causes the display device 12 to display question A2. The operation control unit 151 causes the communication device 13 to send reply A3 to the server 20. The operation control unit 151 may be composed of circuits such as a DSP (Digital Signal Processor), ASIC (Application Specific Integrated Circuit), or FPGA (Field Programmable Gate Array).

[0034] 1-3: Server 20 Figure 4 shows an example of a server 20. Server 20 is an example of a question determination device. Server 20 includes a communication device 21, a storage device 22, a processing device 23, and a bus 24.

[0035] Bus 24 is a wiring system for transmitting information. Bus 24 connects the communication device 21, the storage device 22, and the processing device 23. Bus 24 may consist of a single bus, or it may consist of various buses provided between the devices.

[0036] The communication device 21 can communicate with the terminal 10 via the communication network NW. The communication device 21 may also communicate with the terminal 10 without using the communication network NW. The communication device 21 can communicate with the generated AI 30a via the communication network NW. The communication device 21 may also communicate with the generated AI 30a without using the communication network NW.

[0037] The storage device 22 is a recording medium that can be read by the processing device 23. The storage device 22 includes at least one memory. The storage device 22 includes, for example, non-volatile memory and volatile memory. The storage device 22 stores the program PG2 and the answer candidate information 22a. The program PG2 includes at least one instruction.

[0038] The answer candidate information 22a shows multiple manuals C. These manuals C are classified into multiple answer candidate groups according to the purpose of the inquiry. For the sake of simplicity, the following example describes a case where multiple answer candidate groups are composed of answer candidate group C1 and answer candidate group C2. Note that the multiple answer candidate groups may include three or more answer candidate groups.

[0039] Figure 5 shows an example of a candidate answer group C1. A purpose label L1 is attached to the candidate answer group C1. The purpose label L1 indicates the purpose of the inquiry to which the candidate answer group C1 is classified. The purpose of the inquiry indicated by the purpose label L1 is "New application for service E1 / Change of price plan". The candidate answer group C1 includes manuals D1-1, D1-2, and D1-3. Manual D1-1 is a manual for users who want to sign up for a new contract for service E1. Manual D1-2 is a manual for users who want to change from another service to service E1. Manual D1-3 is a manual for users who want to change from another company's service to service E1. The number of manuals included in the candidate answer group C1 is not limited to three; it may be two, or four or more. The candidate answer group C1 may consist of only one manual.

[0040] Figure 6 shows an example of answer candidate group C2. Answer candidate group C2 is assigned a purpose label L2. The purpose label L2 indicates the purpose of the query to which answer candidate group C2 is classified. The purpose label L2 indicates "Service E1 in general" as the purpose of the query. Answer candidate group C2 includes manuals D2-1, D2-2, and D2-3. Manual D2-1 is a manual for users who want to check the overview of Service E1. Manual D2-2 is a manual for users who want to know the terms and conditions of service E1. Manual D2-3 is a manual for users who want to check the usage fees for Service E1. The number of manuals included in answer candidate group C2 is not limited to 3; it may be 2, or 4 or more. Answer candidate group C2 may consist of only one manual. The number of manuals included in answer candidate group C2 may be the same as, or different from, the number of manuals included in answer candidate group C1.

[0041] In Figure 4, the processing unit 23 is another example of a question determination device. The processing unit 23 responds to inquiry A1 by using answer candidate information 22a. The processing unit 23 includes at least one CPU. The processing unit 23 reads program PG2 from storage device 22. By executing program PG2, the processing unit 23 functions as multiple elements. These multiple elements include an inquiry reception unit 231, a purpose determination unit 232, a question management unit 233, a question presentation unit 234, a response reception unit 235, a dialogue integration unit 236, and a purpose output unit 237. Each of these multiple elements may be composed of a circuit such as a DSP, ASIC, or FPGA.

[0042] Figure 7 shows an example of the functional configuration of the processing unit 23. Each of the various elements of the processing unit 23 will be described below.

[0043] 1-3-1: Inquiry Reception Unit 231 The inquiry reception unit 231 receives inquiry A1. For example, the inquiry reception unit 231 receives inquiry A1 via the communication device 21.

[0044] 1-3-2: Purpose determination unit 232 The purpose determination unit 232 determines the purpose of the inquiry A1. For example, the purpose determination unit 232 determines the purpose of the inquiry A1 based on the inquiry A1. Also, the purpose determination unit 232 determines the purpose of the inquiry A1 based on the inquiry A1, the question A2, and the answer A3. For example, the purpose determination unit 232 determines the purpose of the inquiry A1 based on the updated inquiry A4 based on the inquiry A1, the question A2, and the answer A3.

[0045] The purpose determination unit 232 includes a purpose classification unit 232a and a purpose classification determination unit 232b. Note that the purpose determination unit 232 is not limited to a configuration including the purpose classification unit 232a and the purpose classification determination unit 232b. For example, the purpose determination unit 232 may include a configuration obtained by integrating the purpose classification unit 232a and the purpose classification determination unit 232b.

[0046] When the purpose classification unit 232a receives an inquiry such as the inquiry A1, it outputs at least one purpose label. The at least one purpose label is at least one label indicating a plurality of candidates for the purpose of the inquiry.

[0047] For example, the purpose classification unit 232a inputs a prompt indicating an instruction to generate at least one purpose label based on the inquiry to the generation AI 30a. The purpose classification unit 232a acquires at least one purpose label generated by the generation AI 30a in response to the prompt indicating an instruction to generate at least one purpose label based on the inquiry. The purpose classification unit 232a outputs at least one purpose label acquired from the generation AI 30a.

[0048] FIG. 8 is a diagram showing an example of a prompt F1 indicating an instruction to generate at least one purpose label A5 based on the inquiry A1. In the prompt F1, the inquiry A1 "Tell me about service E1" is described in the inquiry column.

[0049] FIG. 9 is a diagram showing an example of a prompt F2 indicating an instruction to generate at least one purpose label A7 based on the updated inquiry A4. In the prompt F2, the updated inquiry A4 "Please tell me about a new contract for service E1" is described in the inquiry column.

[0050] Figure 10 shows an example of at least one objective label A5 output in response to inquiry A1, "Tell me about service E1". The at least one objective label A5 is generated by generation AI 30a in response to prompt F1 corresponding to inquiry A1, for example. The objective classification unit 232a acquires the at least one objective label A5 generated by generation AI 30a. The objective classification unit 232a outputs the at least one objective label A5 acquired from generation AI 30a. The at least one objective label A5 includes objective label A51 and objective label A52.

[0051] Objective label A51 indicates objective label L1 attached to answer candidate group C1. Objective label L1 indicates "New application for service E1 / Change of price plan". Objective label A51 is associated with output probability D51. Output probability D51 is a value that indicates the probability that the objective candidate for inquiry A1 indicated by objective label A51 is correct. Output probability D51 may also be called the score of objective label A51.

[0052] Objective label A52 indicates objective label L2, which is attached to the candidate answer group C2. Objective label L2 indicates "Service E1 in general". Objective label A52 is associated with output probability D52. Output probability D52 is a value that indicates the probability that the candidate objective of query A1 indicated by objective label A52 is correct. Output probability D52 may also be called the score of objective label A52.

[0053] Figure 11 shows an example of at least one purpose label A7 that is output in response to renewal inquiry A4, which is "Please tell me about a new contract for service E1". The at least one purpose label A7 is generated by the generation AI 30a in response to prompt F2 corresponding to renewal inquiry A4, for example. The purpose classification unit 232a obtains the at least one purpose label A7 generated by the generation AI 30a. The purpose classification unit 232a outputs the at least one purpose label A7 obtained from the generation AI 30a. The at least one purpose label A7 includes purpose label A71 and purpose label A72.

[0054] Objective label A71 indicates objective label L1, which is attached to the answer candidate group C1. Objective label L1 indicates "New application for service E1 / Change of price plan". Objective label A71 is associated with output probability D71. Output probability D71 is a value that indicates the probability that the candidate objective of inquiry A1 indicated by objective label A71 is correct. Output probability D71 may also be called the score of objective label A71.

[0055] Objective label A72 indicates objective label L2, which is attached to the candidate answer group C2. Objective label L2 indicates "Service E1 in general". Objective label A72 is associated with output probability D72. Output probability D72 is a value that indicates the probability that the candidate objective of query A1 indicated by objective label A72 is correct. Output probability D72 may also be called the score of objective label A72.

[0056] In Figure 7, the purpose classification determination unit 232b receives at least one purpose label from the purpose classification unit 232a. Based on the at least one purpose label, the purpose classification determination unit 232b determines whether the purpose classification has been completed.

[0057] For example, if at least one of the target labels has an output probability that exceeds the threshold TH1, the target classification determination unit 232b determines that the target classification is complete. In this case, the target classification determination unit 232b determines the candidate target indicated by the target label with an output probability that exceeds the threshold TH1 as the target of the query.

[0058] On the other hand, if there is no target label with an output probability exceeding the threshold TH1 attached to at least one target label, the target classification determination unit 232b determines that the target classification is incomplete.

[0059] The objective classification determination unit 232b outputs the result A6 of the determination performed by the objective classification determination unit 232b.

[0060] As described above, the objective classification unit 232a outputs at least one objective label by using the generated AI 30a. However, the objective classification unit 232a may output at least one objective label without using the generated AI 30a.

[0061] For example, the objective classification unit 232a may output at least one objective label using a first trained model instead of the generating AI 30a. The first trained model may be included in the server 20, or it may be included in an external device different from the server 20. The server 20 is capable of communicating with the external device containing the first trained model. The first trained model is a model that has learned the relationship between queries and output information. The output information indicates at least one candidate for the query's objective and at least one score. At least one score corresponds one-to-one with at least one candidate for the query's objective. One of the scores included in the at least one score is a value that indicates the probability that the corresponding candidate is correct.

[0062] The objective classification unit 232a inputs the query to the first trained model. The objective classification unit 232a obtains the output information output from the first trained model in response to the query. The objective classification unit 232a outputs the output information obtained from the trained model as at least one objective label. In this case, if there is an objective label among the at least one objective labels that has a score exceeding the threshold TH1, the objective classification determination unit 232b determines that the classification of the objective is complete. The objective classification determination unit 232b determines the candidate objectives indicated by the objective labels that have a score exceeding the threshold TH1 as the objectives of the query. On the other hand, if there is no objective label among the at least one objective label that has a score exceeding the threshold TH1, the objective classification determination unit 232b determines that the classification of the objective is incomplete.

[0063] 1-3-3: Question Management Unit 233 The Question Management Unit 233 manages the output of Question A2 based on the result A6 of the determination made by the Purpose Classification Determination Unit 232b.

[0064] If the result A6 of the determination by the objective classification determination unit 232b indicates that the objective classification is complete, the question management unit 233 does not output question A2.

[0065] If the result A6 of the determination by the objective classification determination unit 232b indicates that the objective classification is incomplete, the question management unit 233 outputs question A2.

[0066] An example of Question A2 is Question A21, which inquires about the purpose of the inquiry. Another example of Question A2 is Question A22, which confirms the purpose of the inquiry. An example of an inquiry is Inquiry A1. Another example of an inquiry is Update Inquiry A4.

[0067] Figure 12 shows an example of question A21, which inquires about the purpose of the inquiry. Question A21 shown in Figure 12 is the same as question A2 shown in Figure 2.

[0068] Figure 13 shows an example of Question A22, which confirms the purpose of the inquiry. Question A22 is, "Is your inquiry about the usage fees for Service E1?"

[0069] If the result A6 of the determination by the objective classification determination unit 232b indicates that the objective classification is incomplete, the question management unit 233 outputs either question A21 or question A22 based on the relationship between the inquiry and the candidate answer information 22a. An example of the question management unit 233 will be described later.

[0070] 1-3-4: Question Presentation Unit 234 In Figure 7, the question presentation unit 234 presents question A2 to user U1 using the display device 12 of terminal 10. For example, the question presentation unit 234 causes the display device 12 to display question A2 by providing terminal 10 with a message indicating question A2. The question presentation unit 234 presents question A2 to user U1 by displaying question A2 on the display device 12.

[0071] 1-3-5: Reply Reception Unit 235 The reply reception unit 235 receives reply A3. For example, the reply reception unit 235 receives reply A3 via the communication device 21.

[0072] 1-3-6: Dialogue Integration Unit 236 The dialogue integration unit 236 integrates the dialogue between user U1 and server 20. The dialogue integration unit 236 determines update inquiry A4 based on inquiry A1, question A2, and response A3.

[0073] For example, the dialogue integration unit 236 inputs a prompt F3 to the generation AI 30a, which indicates an instruction to generate a sentence based on inquiry A1, question A2, and response A3. The dialogue integration unit 236 retrieves the sentence generated by the generation AI 30a in response to prompt F3. The dialogue integration unit 236 determines the sentence retrieved from the generation AI 30a as update inquiry A4.

[0074] Figure 14 shows an example of prompt F3, which indicates an instruction to generate a sentence based on inquiry A1, question A2, and response A3. In prompt F3, inquiry A1, question A2, and response A3 are written in the dialogue history field.

[0075] As described above, the dialogue integration unit 236 shown in Figure 7 outputs update query A4 by using the generated AI 30a. However, the dialogue integration unit 236 may output update query A4 without using the generated AI 30a. For example, the dialogue integration unit 236 may generate update query A4 by combining query A1, question A2, and response A3.

[0076] 1-3-7: Objective Output Unit 237 The objective output unit 237 outputs the objective of inquiry A1 determined by the objective determination unit 232.

[0077] In a situation where at least one objective label A5 is output, if the result A6 of the determination by the objective classification determination unit 232b indicates that the classification of the objective is complete, the objective output unit 237 outputs the objective label with the highest output probability among the at least one objective label A5 as the objective of query A1.

[0078] In a situation where at least one objective label A7 is output, if the result A6 of the determination by the objective classification determination unit 232b indicates that the classification of the objective is complete, the objective output unit 237 outputs the objective label with the highest output probability among the at least one objective label A7 as the objective of query A1.

[0079] The purpose of inquiry A1 is to determine which manual C is appropriate to the purpose of inquiry A1 from among several manuals C. However, the purpose of inquiry A1 does not necessarily have to be used to determine which manual C is appropriate to the purpose of inquiry A1. For example, the purpose of inquiry A1 may be communicated to the operator corresponding to user U1.

[0080] 1-4: Question Management Unit 233 Figure 15 shows an example of the question management unit 233 shown in Figure 7.

[0081] Figure 15 shows inquiry A1. Note that if the dialogue integration unit 236 shown in Figure 7 outputs update inquiry A4, then in Figure 15, update inquiry A4 will be used instead of inquiry A1.

[0082] Furthermore, Figure 15 shows at least one objective label A5 corresponding to query A1. Note that if the objective classification unit 232a shown in Figure 7 outputs at least one objective label A7 corresponding to update query A4, then in Figure 15, at least one objective label A7 corresponding to update query A4 is used instead of at least one objective label A5 corresponding to query A1.

[0083] The question management unit 233 includes a summary unit 1, a candidate answer determination unit 2, a category determination unit 3, a related category determination unit 4, and a question determination unit 5.

[0084] 1-4-1: Summarization Unit 1 When Summarization Unit 1 receives query A1, it outputs a summary A8 of query A1. The summary A8 of query A1 shows substantially the same content as what query A1 means. The summary A8 of query A1 does not include any extraneous information that query A1 may have.

[0085] The summarization unit 1 obtains a summary A8 of query A1 by using the generation AI 30a. For example, the summarization unit 1 inputs a prompt F4 to the generation AI 30a, which indicates an instruction to generate a summary of query A1. The summarization unit 1 obtains the summary generated by the generation AI 30a in response to prompt F4 as summary A8.

[0086] Figure 16 shows an example of prompt F4, which indicates an instruction to generate a summary of inquiry A1. In prompt F4, inquiry A1, "Tell me about service E1," is written in the inquiry field.

[0087] Figure 17 shows an example of summary A8 generated by generation AI 30a in response to prompt F4. Summary A8 indicates "Inquiry for service E1".

[0088] As described above, the summarization unit 1 shown in Figure 15 obtains query A1 by using the generated AI 30a. However, the summarization unit 1 may obtain query A1 without using the generated AI 30a.

[0089] For example, the summarization unit 1 may use a second trained model instead of the generating AI 30a to summarize query A1. The second trained model may be included in the server 20 or in an external device. The server 20 can communicate with the external device, which includes the second trained model. The second trained model is a model that has learned the relationship between a sentence and a summary of that sentence. The summarization unit 1 inputs query A1 to the second trained model. The summarization unit 1 obtains the summary output from the second trained model in response to query A1 as summary A8.

[0090] 1-4-2: Answer Candidate Determination Unit 2 The Answer Candidate Determination Unit 2 determines at least one manual A9 from among multiple manuals C based on inquiry A1. Multiple manuals C are examples of multiple answer candidates. At least one manual A9 is an example of at least one answer candidate related to inquiry A1 from user U1.

[0091] For example, if the result A6 of the determination by the purpose classification determination unit 232b indicates that the purpose classification is incomplete, the answer candidate determination unit 2 determines at least one manual A9 based on at least one purpose label A5 and inquiry A1.

[0092] First, the answer candidate determination unit 2 selects at least one group of answer candidates from among the multiple groups of answer candidates stored in the storage device 22 as at least one selection candidate.

[0093] At least one selection candidate is comprised of, for example, a group of answer candidates to which at least one of the answer candidate groups stored in the storage device 22 has an answer label included in the answer label A5.

[0094] If at least one objective label A5 includes multiple objective labels, at least one selection candidate consists of a group of answer candidates to which any of the multiple objective labels are attached.

[0095] Furthermore, if at least one objective label A5 includes multiple objective labels, at least one selection candidate may consist of a first response candidate group to which the objective label having the highest output probability among the multiple objective labels A5 is attached.

[0096] Furthermore, if at least one objective label A5 includes multiple objective labels, at least one selection candidate may consist of a first group of candidate responses and a second group of candidate responses to which the objective label having the second highest output probability among the multiple objective labels A5 is attached.

[0097] If at least one objective label A5 is comprised of one objective label, then at least one selection candidate is comprised of a group of answer candidates to which one objective label is attached.

[0098] The answer candidate determination unit 2 searches for at least one manual A9 related to inquiry A1 from among at least one manual included in at least one selection candidate. The search methods used by the answer candidate determination unit 2 include, for example, full-text search, vector search, full-text / vector hybrid search, search using full-text search and semantic ranking in combination, or search using full-text / vector hybrid search and semantic ranking in combination.

[0099] The answer candidate determination unit 2 outputs reference information indicating at least one manual A9 to the category determination unit 3.

[0100] 1-4-3: Category determination unit 3 The category determination unit 3 determines at least one category A10 to which at least one manual A9 belongs, based on at least one manual A9.

[0101] For example, the category determination unit 3 inputs a prompt F5 to the generation AI 30a indicating an instruction to generate at least one category A10 to which at least one manual A9 belongs. The category determination unit 3 obtains at least one category generated by the generation AI 30a in response to prompt F5. The category determination unit 3 determines the at least one category obtained from the generation AI 30a as at least one category A10.

[0102] Figure 18 shows an example of a prompt F5 that indicates an instruction to generate at least one category A10 to which at least one manual A9 belongs. In the prompt F5 shown in Figure 18, the titles of multiple manuals C are described as an example of at least one manual A9. However, the example of at least one manual A9 in prompt F5 is not limited to the titles of multiple manuals C. For example, the example of at least one manual A9 in prompt F5 may be multiple manuals C.

[0103] Figure 19 shows an example of at least one category generated by the generating AI 30a in response to prompt F5. In Figure 19, the at least one category generated by the generating AI 30a in response to prompt F5 is shown as at least one category A10.

[0104] As described above, the category determination unit 3 shown in Figure 15 determines at least one category A10 by using the generated AI 30a. However, the category determination unit 3 may determine at least one category A10 without using the generated AI 30a.

[0105] For example, the category determination unit 3 may determine at least one category A10 using a second trained model instead of the generating AI 30a. The category determination unit 3 inputs at least one manual A9 into the second trained model. If at least one manual A9 includes multiple manuals, the category determination unit 3 inputs the multiple manuals individually into the second trained model. The category determination unit 3 may determine at least one summary output from the second trained model in accordance with at least one manual A9 as at least one category A10.

[0106] 1-4-4: Related Category Determination Unit 4 The related category determination unit 4 determines at least one related category A11 related to query A1 from at least one category A10.

[0107] For example, the related category determination unit 4 determines at least one related category A11 related to query A1 from among the at least one category A10, based on the relationship between the summary A8 of query A1 and at least one category A10.

[0108] For example, the related category determination unit 4 determines at least one related category A11 based on relatedness information A12, which indicates the degree of relatedness between at least one category A10 and the summary A8 of query A1. The degree of relatedness between at least one category A10 and the summary A8 of query A1 is an example of the degree of relatedness between at least one category A10 and query A1.

[0109] The related category determination unit 4 may determine at least one related category A11 based on relatedness information indicating the degree of relevance between at least one category A10 and query A1, instead of the relatedness information A12. In this case, the summarization unit 1 may be omitted.

[0110] Furthermore, the configuration including the summarization unit 1 has the following effects: The summary A8 of query A1 determined by the summarization unit 1 does not contain any extraneous information that query A1 may have. Therefore, the configuration in which the related category determination unit 4 determines at least one related category A11 using the summary A8 of query A1 has the effect of determining at least one related category A11 with higher accuracy than the configuration in which the related category determination unit 4 determines at least one related category A11 using query A1 instead of the summary A8 of query A1. This effect is particularly noticeable when query A1 is redundant.

[0111] The relevance information A12, which indicates the degree of relevance between at least one category A10 and the summary A8 of query A1, is generated by the generation AI 30a, for example, based on at least one category A10 and query A1.

[0112] The related category determination unit 4 includes a relevance determination unit 6 and a question determination unit 7. However, the related category determination unit 4 is not limited to a configuration that includes a relevance determination unit 6 and a question determination unit 7. For example, the related category determination unit 4 may include a configuration obtained by integrating the relevance determination unit 6 and the question determination unit 7.

[0113] 1-4-4-1: Relevance Determination Unit 6 The relevance determination unit 6 determines relevance information A12 that indicates the degree of relevance between at least one category A10 and the summary A8 of query A1. For example, the relevance determination unit 6 inputs a prompt F6 to the generation AI 30a indicating an instruction to generate relevance information A12 based on at least one category A10 and the summary A8 of query A1. The relevance determination unit 6 obtains the relevance information generated by the generation AI 30a in response to prompt F6. The relevance determination unit 6 determines the relevance information obtained from the generation AI 30a as relevance information A12.

[0114] Figure 20 shows an example of prompt F6, which instructs the system to generate relevance information A12 based on at least one category A10 and a summary A8 of query A1. Prompt F6 instructs the system to output a score indicating the degree of relevance between at least one category A10 and a summary A8 of query A1.

[0115] Figure 21 shows an example of relevance information generated by the generating AI 30a in response to prompt F6. In Figure 21, the relevance information generated by the generating AI 30a in response to prompt F6 is shown as relevance information A12. In Figure 21, each of the multiple categories is assigned a relevance score indicated by a numerical value. For example, the category "Regarding new contracts for service E1" is assigned a relevance score indicated by "3".

[0116] As described above, the relevance determination unit 6 shown in Figure 15 determines the relevance information A12 by using the generated AI 30a. However, the relevance determination unit 6 may determine the relevance information A12 without using the generated AI 30a.

[0117] For example, the relevance determination unit 6 converts at least one category A10 into at least one vector. The relevance determination unit 6 converts the summary A8 of query A1 into a vector. Subsequently, the relevance determination unit 6 calculates the cosine similarity between at least one vector based on at least one category A10 and the vector based on the summary A8 of query A1 as the degree of relevance between at least one category A10 and the summary A8 of query A1. Subsequently, the relevance determination unit 6 determines information indicating the cosine similarity between at least one category A10 as relevance information A12.

[0118] If at least one category A10 has multiple categories, the relevance determination unit 6 converts each of the multiple categories into a vector. The relevance determination unit 6 calculates the cosine similarity between each of the multiple category vectors and the vector based on the summary A8 of query A1 as the degree of relevance between each of the multiple categories and the summary A8 of query A1. The relevance determination unit 6 determines information indicating each of the multiple categories and the corresponding cosine similarity as relevance information A12.

[0119] 1-4-4-2: Question Determination Unit 7 The Question Determination Unit 7 determines, based on the relevance information A12, whether or not a question is necessary to inquire about the purpose of the inquiry. Hereinafter, a question inquiring about the purpose of the inquiry will be referred to as a "purpose inquiry question".

[0120] The question determination unit 7 first determines at least one related category A11 related to inquiry A1 from at least one category A10 based on the relevance information A12. For example, the question determination unit 7 determines at least one related category A11 from at least one category A10 based on the threshold TH2 and the relevance information A12.

[0121] The threshold TH2 is, for example, 2.5. However, the threshold TH2 is not limited to 2.5. For example, the threshold TH2 may be greater than 2.5. Alternatively, the threshold TH2 may be a value within the range of 1.5 or more and less than 2.5.

[0122] For example, if at least one category A10 includes at least one category having a relevance of threshold TH2 or higher, the question determination unit 7 determines that at least one category having a relevance of threshold TH2 or higher is at least one related category A11.

[0123] If at least one category A10 is comprised of one category that does not have a relevance of threshold TH2 or higher, the question determination unit 7 determines that one category as at least one related category A11.

[0124] If at least one category A10 is composed of multiple categories that do not have a relevance of threshold TH2 or higher, the question determination unit 7 determines that the first category with the highest relevance among the multiple categories and the second category with the second highest relevance among the multiple categories are at least one related category A11.

[0125] If at least one related category A11 includes multiple related categories, the question determination unit 7 determines that a purpose inquiry question is necessary. If the question determination unit 7 determines that a purpose inquiry question is necessary, it provides at least one related category A11 to the purpose inquiry determination unit 8 included in the question determination unit 5.

[0126] If at least one related category A11 is comprised of one related category, the question determination unit 7 determines that a purpose inquiry question is unnecessary. When the question determination unit 7 determines that a purpose inquiry question is unnecessary, it provides at least one related category A11 to the purpose confirmation determination unit 9 included in the question determination unit 5.

[0127] 1-4-5: Question Determination Unit 5 The Question Determination Unit 5 determines Question A2 concerning the purpose of Inquiry A1 by using at least one related category A11.

[0128] If at least one related category A11 includes multiple related categories related to inquiry A1, then question A2 is question A21, which inquires about the purpose of inquiry A1. Question A21 prompts user U1 to select one related category related to the purpose of inquiry A1 from among the multiple related categories related to inquiry A1.

[0129] If at least one related category A11 is related to inquiry A1, then question A2 is question A22, which confirms the purpose of the inquiry. Question A22 prompts user U1 to confirm that one related category related to inquiry A1 is related to the purpose of inquiry A1.

[0130] The question determination unit 5 includes a purpose inquiry determination unit 8 and a purpose confirmation determination unit 9.

[0131] 1-4-5-1: Purpose Inquiry Determination Unit 8 The purpose inquiry determination unit 8 determines question A21 by using multiple related categories that are included in at least one related category A11.

[0132] For example, the objective query determination unit 8 inputs a prompt F7 to the generation AI 30a, indicating an instruction to generate question A21 based on multiple related categories. The objective query determination unit 8 retrieves the question generated by the generation AI 30a in response to prompt F7. The objective query determination unit 8 determines the question retrieved from the generation AI 30a as question A21.

[0133] Figure 22 shows an example of prompt F7, which indicates instructions for generating question A21 based on multiple related categories. In prompt F7 shown in Figure 22, two related categories are described, consisting of the category "Regarding a new contract for service E1" and the category "Regarding an overview of service E2".

[0134] In response to prompt F7, the generating AI 30a generates, for example, question A21 shown in Figure 12.

[0135] As described above, the objective query determination unit 8 shown in Figure 15 determines question A21 by using the generated AI 30a. However, the objective query determination unit 8 may determine question A21 without using the generated AI 30a.

[0136] For example, the objective inquiry determination unit 8 may determine question A21 using at least one related category A11 and a template.

[0137] Figure 23 shows an example of template T1 used to generate question A21. Multiple related categories are added to template T1 shown in Figure 23.

[0138] Figure 24 shows another example of Question A21 generated based on Template T1, which has two related categories added: "Regarding a New Contract for Service E1" and "Regarding an Overview of Service E2".

[0139] Question A21 is provided to the question presentation unit 234 and the dialogue integration unit 236, as shown in Figure 7.

[0140] 1-4-5-2: Objective Confirmation and Decision Unit 9 In Figure 15, the objective confirmation and decision unit 9 determines question A22 by using one related category that constitutes at least one related category A11.

[0141] For example, the objective confirmation and decision unit 9 inputs a prompt F8 to the generation AI 30a, indicating an instruction to generate question A22 based on one related category. The objective confirmation and decision unit 9 retrieves the question generated by the generation AI 30a in response to prompt F8. The objective confirmation and decision unit 9 determines the question retrieved from the generation AI 30a as question A22.

[0142] Figure 25 shows an example of prompt F8, which indicates instructions to generate question A22 based on one relevant category. In prompt F8 shown in Figure 25, one relevant category is described as "Regarding the usage fees for service E1".

[0143] In response to prompt F8, the generating AI 30a generates, for example, question A22 shown in Figure 13.

[0144] As described above, the objective confirmation and decision unit 9 shown in Figure 15 determines question A22 by using the generated AI 30a. However, the objective confirmation and decision unit 9 may determine question A22 without using the generated AI 30a.

[0145] For example, the objective confirmation and determination unit 9 may determine question A22 using at least one related category A11 and a template.

[0146] Figure 26 shows an example of template T2 used to generate question A22. One related category is added to template T2 shown in Figure 26.

[0147] Figure 27 shows another example of question A22 generated based on template T2 with the addition of one related category, "Regarding the usage fees for service E1."

[0148] Question A22 is provided to the question presentation unit 234 and the dialogue integration unit 236, as shown in Figure 7.

[0149] 1-5: Diagrams 28 and 29 are diagrams illustrating the operation of the server 20.

[0150] In step S101, the inquiry reception unit 231 receives inquiry A1 from user U1.

[0151] For example, the inquiry reception unit 231 receives inquiry A1 from user U1 via the communication device 21 from terminal 10.

[0152] Next, in step S102, the purpose classification unit 232a outputs at least one purpose label A5 based on inquiry A1.

[0153] For example, the objective classification unit 232a inputs a prompt F1 to the generation AI 30a indicating an instruction to generate at least one objective label A5 based on the inquiry A1. The objective classification unit 232a obtains at least one objective label A5 generated by the generation AI 30a in response to prompt F1. The objective classification unit 232a outputs the at least one objective label A5 obtained from the generation AI 30a.

[0154] Next, in step S103, the purpose classification determination unit 232b determines whether the purpose classification has been completed based on at least one purpose label A5.

[0155] For example, if at least one objective label A5 has an output probability that exceeds the threshold TH1, the objective classification determination unit 232b determines that the classification of objectives is complete. In this case, the objective classification determination unit 232b determines the candidate objectives indicated by the objective labels with output probabilities exceeding the threshold TH1 as the objectives of the query.

[0156] If there is no target label A5 that has an output probability exceeding the threshold TH1, the target classification determination unit 232b determines that the target classification is incomplete.

[0157] The objective classification determination unit 232b outputs the result A6 of the determination performed by the objective classification determination unit 232b.

[0158] If the objective classification determination unit 232b determines in step S103 that the objective classification has been completed, step S114 is executed.

[0159] In step S114, the objective output unit 237 outputs the objective label with the highest output probability among at least one objective label A5 as the objective of query A1.

[0160] On the other hand, if the objective classification determination unit 232b determines in step S103 that the objective classification has not been completed, step S104 is executed.

[0161] In step S104, the answer candidate determination unit 2 determines at least one manual A9 from among multiple manuals C based on inquiry A1.

[0162] For example, the answer candidate determination unit 2 first determines at least one group of answer candidates from among the multiple groups of answer candidates stored in the storage device 22 as at least one selected candidate. The at least one selected candidate is, for example, composed of a group of answer candidates to which at least one of the target labels A5 is attached, from among the multiple groups of answer candidates stored in the storage device 22.

[0163] Next, the answer candidate determination unit 2 searches for at least one manual A9 related to inquiry A1 from among at least one manual included in at least one selection candidate. The search methods used by the answer candidate determination unit 2 include, for example, full-text search, vector search, full-text / vector hybrid search, search using full-text search and semantic ranking in combination, or search using full-text / vector hybrid search and semantic ranking in combination.

[0164] Next, in step S105, the category determination unit 3 determines at least one category A10 to which at least one manual A9 belongs, based on at least one manual A9.

[0165] For example, the category determination unit 3 inputs a prompt F5 to the generation AI 30a indicating an instruction to generate at least one category A10 to which at least one manual A9 belongs. The category determination unit 3 obtains at least one category generated by the generation AI 30a in response to prompt F5. The category determination unit 3 determines the at least one category obtained from the generation AI 30a as at least one category A10.

[0166] Next, in step S106, the summarization unit 1 outputs a summary A8 of inquiry A1.

[0167] For example, the summarization unit 1 inputs a prompt F4 to the generation AI 30a, indicating an instruction to generate a summary of inquiry A1. The summarization unit 1 then obtains the summary generated by the generation AI 30a in response to prompt F4 as summary A8.

[0168] Step S106 may be executed during the period from the end of step S103 to the start of step S104. Alternatively, step S106 may be executed during the period from the end of step S104 to the start of step S105. Step S106 may be executed in parallel with step S104. Step S106 may be executed in parallel with step S105.

[0169] Next, in step S107, the relevance determination unit 6 determines relevance information A12 that indicates the degree of relevance between at least one category A10 and the summary A8 of inquiry A1.

[0170] For example, the relevance determination unit 6 inputs a prompt F6 to the generation AI 30a, indicating an instruction to generate relevance information A12. The relevance determination unit 6 acquires the relevance information generated by the generation AI 30a in response to prompt F6. The relevance determination unit 6 determines the relevance information acquired from the generation AI 30a as relevance information A12.

[0171] Next, in step S108, the question determination unit 7 determines whether a purpose inquiry question is necessary based on the relevance information A12.

[0172] The question determination unit 7 first determines, based on the relevance information A12, at least one related category A11 related to inquiry A1 from at least one category A10.

[0173] For example, the question determination unit 7 determines at least one related category A11 from at least one category A10 based on the threshold TH2 and the relatedness information A12.

[0174] For example, if at least one category A10 includes at least one category with a relevance of threshold TH2 or higher, the question determination unit 7 determines that at least one category with a relevance of threshold TH2 or higher is at least one related category A11.

[0175] If at least one category A10 is comprised of one category that does not have a relevance of threshold TH2 or higher, the question determination unit 7 determines that one category as at least one related category A11.

[0176] If at least one category A10 is composed of multiple categories that do not have a relevance of threshold TH2 or higher, the question determination unit 7 determines that the first category with the highest relevance among the multiple categories and the second category with the second highest relevance among the multiple categories are at least one related category A11.

[0177] If at least one related category A11 includes multiple related categories, the question determination unit 7 determines that a purpose query question is necessary. If the question determination unit 7 determines that a purpose query question is necessary, it provides at least one related category A11 to the purpose query determination unit 8.

[0178] If at least one related category A11 is comprised of one related category, the question determination unit 7 determines that a purpose inquiry question is unnecessary. When the question determination unit 7 determines that a purpose inquiry question is unnecessary, it provides at least one related category A11 to the purpose confirmation determination unit 9.

[0179] If the question determination unit 7 determines in step S108 that a purpose inquiry question is necessary, then step S109 is executed.

[0180] In step S109, the objective query determination unit 8 determines question A21, which is the objective query question, by using multiple related categories that are included in at least one related category A11.

[0181] For example, the objective query determination unit 8 inputs prompt F7 to the generation AI 30a, indicating an instruction to generate question A21. The objective query determination unit 8 retrieves the question generated by the generation AI 30a in response to prompt F7. The objective query determination unit 8 determines the question retrieved from the generation AI 30a as question A21.

[0182] On the other hand, if the question determination unit 7 determines in step S108 that a purpose inquiry question is not necessary, step S110 is executed.

[0183] In step S110, the objective confirmation unit 9 determines question A22, which is a question to confirm the purpose of the inquiry, by using one related category that constitutes at least one related category A11. Hereinafter, questions to confirm the purpose of the inquiry will be referred to as "objective confirmation questions." Question A22 is an example of an objective confirmation question.

[0184] For example, the objective confirmation and decision unit 9 inputs a prompt F8 to the generation AI 30a, indicating an instruction to generate question A22. The objective confirmation and decision unit 9 retrieves the question generated by the generation AI 30a in response to prompt F8. The objective confirmation and decision unit 9 determines the question retrieved from the generation AI 30a as question A22.

[0185] Next, in step S111, the question presentation unit 234 presents question A2 to user U1 using the display device 12 of terminal 10.

[0186] For example, if the objective inquiry determination unit 8 determines question A22, which is an example of question A2, the question presentation unit 234 presents question A21 to user U1 using the display device 12 of terminal 10.

[0187] On the other hand, if the objective confirmation and decision unit 9 determines question A22, which is an example of question A2, the question presentation unit 234 presents question A22 to user U1 using the display device 12 of terminal 10.

[0188] User U1 inputs response A3 to question A2 presented on the display device 12 of terminal 10 into terminal 10. Terminal 10 provides response A3 to server 20.

[0189] Next, in step S112, the response receiving unit 235 receives response A3 from user U1.

[0190] For example, the response receiving unit 235 receives response A3 from the terminal 10 via the communication device 21.

[0191] Next, in step S113, the dialogue integration unit 236 determines an update inquiry A4 based on inquiry A1, question A2, and response A3.

[0192] For example, the dialogue integration unit 236 inputs a prompt F3 to the generation AI 30a, which indicates an instruction to generate a sentence based on inquiry A1, question A2, and response A3. The dialogue integration unit 236 retrieves the sentence generated by the generation AI 30a in response to prompt F3. The dialogue integration unit 236 determines the sentence retrieved from the generation AI 30a as update inquiry A4.

[0193] Next, the process returns to step S102. The loop from step S102 to step S113 continues until it is determined in step S103 that the desired classification has been completed. Note that in steps S102, S104, and S106, which start after step S113, update query A4 is used instead of query A1. Also, in the new step S113 that starts after step S113, the dialogue integration unit 236 determines a new update query A4 based on query A1, question A2, answer A3, a new question, and a new answer.

[0194] 1-6: Summary of Embodiments The server 20 includes a category determination unit 3, a related category determination unit 4, and a question determination unit 5.

[0195] The category determination unit 3 determines at least one category A10 to which at least one manual A9 belongs, based on at least one manual A9 related to inquiry A1 from user U1. The related category determination unit 4 determines at least one related category A11 related to inquiry A1 from at least one category A10. The question determination unit 5 determines question A2 concerning the purpose of inquiry A1 by using at least one related category A11.

[0196] Therefore, it becomes possible to provide user U1 with question A2 that does not include classification symbols such as "Chapter 1, Section 2". Consequently, the possibility of question A2 becoming a redundant question that includes classification symbols such as "Chapter 1, Section 2" can be reduced.

[0197] In addition to inquiry A1, question A2 is determined using at least one manual A9 that user U1 is unaware of. Therefore, an appropriate question A2 can be determined regardless of user U1's skill level.

[0198] Furthermore, at least one category A10 to which at least one manual A9 belongs succinctly expresses the gist of at least one manual A9. Therefore, even if at least one manual A9 that user U1 is unfamiliar with is used, a question A2 that is easy for user U1 to understand can be determined.

[0199] Furthermore, even if at least one category A10 includes categories unrelated to inquiry A1, the related category determination unit 4 determines at least one related category A11 that does not include categories unrelated to inquiry A1. This reduces the possibility that question A2 will be determined based on categories unrelated to inquiry A1. In addition, the related category determination unit 4 can simultaneously eliminate categories with low relevance and identify related categories related to question A2.

[0200] Furthermore, it eliminates the need to prepare sets of questions and answers in advance. This reduces the workload required to prepare these sets. Also, systems that use sets of questions and answers have the problem of standardized answers. In this embodiment, since sets of questions and answers are not required, it is possible to respond flexibly to inquiry A1.

[0201] When at least one related category A11 includes multiple related categories related to inquiry A1, the question A2 determined by the question determination unit 5 may be a question that prompts user U1 to select one related category related to the purpose of inquiry A1 from among the multiple related categories. In other words, when at least one related category A11 includes multiple related categories related to inquiry A1, question A2 may also be a purpose inquiry question. In this case, user U1 only needs to respond by selecting one related category from among the multiple related categories. Therefore, a question that is easy for user U1 to answer can be used as question A2.

[0202] When at least one related category A11 is one related category related to inquiry A1, the question A2 determined by the question determination unit 5 may be a question prompting user U1 to confirm that one related category is related to the purpose of inquiry A1. In other words, when at least one related category A11 is one related category related to inquiry A1, question A2 may be a purpose confirmation question. In this case, user U1 only needs to respond by confirming one related category. Therefore, the burden on user U1 to respond can be reduced. In addition, it can be prevented from having user U1 select a related category more than necessary.

[0203] Furthermore, the purpose inquiry question and the purpose confirmation question can be switched between each other based on the number of related categories that make up at least one related category A11.

[0204] At least one category A10 may be generated by the generation AI 30a in response to instructions from the category determination unit 3. The instructions from the category determination unit 3 are instructions to generate at least one category A10 based on at least one manual A9. In this case, at least one category A10 can be generated by using the generation AI 30a.

[0205] The related category determination unit 4 may determine at least one related category A11 based on relatedness information A12 indicating the degree of relatedness between at least one category A10 and query A1. In this case, at least one related category A11 can be determined based on the degree of relatedness between at least one category A10 and query A1.

[0206] The relevance information A12 may be generated by the generation AI 30a in response to instructions from the relevance category determination unit 4. The instructions from the relevance category determination unit 4 are instructions to generate the relevance information A12 based on at least one category A10 and query A1. In this case, the relevance information A12 can be generated by using the generation AI 30a.

[0207] The server 20 may further include a question presentation unit 234 and a purpose determination unit 232. The question presentation unit 234 presents question A2 to user U1 using the display device 12 of terminal 10. The purpose determination unit 232 determines the purpose of inquiry A1 based on inquiry A1, question A2, and response A3. In this case, for example, even if the purpose of inquiry A1 cannot be determined from inquiry A1 alone, the purpose of inquiry A1 can be determined based on inquiry A1, question A2, and response A3.

[0208] The purpose determination unit 232 may determine the purpose of inquiry A1 based on the update inquiry A4. The update inquiry A4 is a sentence generated by the generation AI 30a in response to an instruction to generate a sentence based on the response A3, question A2, and inquiry A1. In this case, based on the dialogue between inquiry A1, question A2, and response A3, the update inquiry A4 can include information that is missing from inquiry A1. This improves the accuracy of purpose determination.

[0209] The server 20 may further include a response candidate determination unit 2 that determines at least one manual A9 from among multiple manuals C based on inquiry A1. In this case, at least one manual A9 can be determined from among multiple manuals C based on inquiry A1.

[0210] 2: Modifications The following are examples of modifications to the above-described embodiments. Two or more modifications can be arbitrarily selected from the following examples and combined as appropriate, within the bounds of mutual non-contradictory relationships.

[0211] 2-1: In the first modified embodiment, each of the prompts F1 to F8 is not limited to the example described above and can be changed as appropriate. Each of the prompts F1 to F8 is generated by the processing unit 23 of the server 20.

[0212] 2-2: In the second modified embodiment, the multiple manuals C do not have to be stored in the server 20. For example, they may be stored in an external device different from the server 20. In this case, the server 20 reads the multiple manuals C from the external device.

[0213] 3. Other (1) Each function illustrated in Figures 3, 4, 7, or 15 can be implemented by any combination of hardware and software. The method of implementing each function is not particularly limited. Each function may be implemented using one physically or logically coupled device, or it may be implemented using a device configured by directly or indirectly connecting two or more physically or logically separated devices (for example, using wired or wireless connections). Each function may be implemented by combining the above one device or the above multiple devices with software.

[0214] (2) In this disclosure, the term “apparatus” may be replaced with other terms such as circuit, device or unit.

[0215] (3) In each of the embodiments and the first to second modifications, the storage device 14 and the storage device 22 may consist of at least one of the following: an optical disc such as a CD-ROM (Compact Disc ROM), a hard disk drive, a flexible disk, a magneto-optical disk (e.g., compact disc, digital multipurpose disc, Blu-ray® disc), a smart card, flash memory (e.g., card, stick, key drive), a floppy® disk, a magnetic strip, etc. The program may also be transmitted from a network via a telecommunications line.

[0216] (4) Each of the embodiments and the first to second modified examples is based on LTE (Long Term Evolution), LTE-A (LTA-Advanced), SUPER 3G, IMT-Advanced, 4G (4th generation mobile communication system), 5G (5th generation mobile communication system), 6th generation mobile communication system (6G), xth generation mobile communication system (xG) (where x is, for example, an integer or decimal), FRA (Future Radio Access), NR (new Radio), New radio access (NX), Future generation radio access (FX), W-CDMA (registered trademark), GSM (registered trademark), CDMA2000, UMB (Ultra Mobile Broadband), IEEE 802.11 (Wi-Fi (registered trademark)), IEEE 802.16 (WiMAX (registered trademark)), IEEE 802.20 may apply to at least one system utilizing UWB (Ultra-WideBand), Bluetooth®, or other appropriate systems, and to next-generation systems extended, modified, created, or defined based thereon. Alternatively, multiple systems may be applied in combination (e.g., a combination of at least one of LTE and LTE-A with 5G).

[0217] (5) The processing procedures, sequences, flowcharts, etc., exemplified in each of the embodiments and the first to second modifications may be rearranged in order, as long as they do not contradict each other. For example, the methods described in this disclosure present various step elements in an exemplary order and are not limited to the specific order presented.

[0218] (6) In each of the embodiments and the first to second modifications, the input and output information may be stored in a specific location (e.g., memory) or managed using a management table. The input and output information may be overwritten, updated, or appended to. The output information may be deleted. The input information may be transmitted to other devices.

[0219] (7) In each of the embodiments and the first to second modifications, the determination may be based on a value represented by one bit (0 or 1), on a Boolean value (true or false), or on a numerical comparison (for example, a comparison with a predetermined value).

[0220] (8) The programs illustrated in each of the embodiments and the first to second modifications should be broadly interpreted to mean instructions, instruction sets, code, code segments, program code, subprograms, software modules, applications, software applications, software packages, routines, subroutines, objects, executable files, execution threads, procedures, or functions, whether they are called software, firmware, middleware, microcode, or hardware description languages ​​or by other names. Furthermore, software, or instructions, etc., may be transmitted or received via a transmission medium. For example, if software is transmitted from a website, server, or other remote source using at least one of wired technology (such as coaxial cable, fiber optic cable, twisted pair, and digital subscriber line (DSL)) and wireless technology (such as infrared, microwave, etc.), at least one of these wired and wireless technologies is included in the definition of a transmission medium.

[0221] (9) The information described in each of the embodiments and the first to second modifications may be represented using any of the following different technologies. For example, the data and information that may be referred to throughout the above description may be represented by voltage, current, electromagnetic waves, magnetic fields, magnetic particles, optical fields, photons, or any combination thereof. Notwithstanding the terms described herein and the terms necessary for understanding this disclosure, terms may be replaced with terms having the same or similar meanings.

[0222] (10) In each of the embodiments and the first to second modifications, the terms “system” and “network” are used interchangeably.

[0223] (11) In each of the embodiments and the first to second modifications, the terminal 10 is, for example, a mobile station. A mobile station may also be referred to by those skilled in the art as a subscriber station, mobile unit, subscriber unit, wireless unit, remote unit, mobile device, wireless device, wireless communication device, remote device, mobile subscriber station, access terminal, mobile terminal, wireless terminal, remote terminal, handset, user agent, mobile client, client, or several other appropriate terms.

[0224] (12) A mobile station may also be called a transmitting device, receiving device, or communication device. A mobile station may also be a device mounted on a mobile body, or the mobile body itself. A mobile body means a movable object. The speed of movement of a mobile body is arbitrary. A mobile body is stoppable. A mobile body includes, but is not limited to, vehicles, transport vehicles, automobiles, motorcycles, bicycles, connected cars, excavators, bulldozers, wheel loaders, dump trucks, forklifts, trains, buses, handcarts, rickshaws, ships and other watercraft, airplanes, rockets, satellites, drones (registered trademark), multicopters, quadcopters, balloons, and things mounted on them. A mobile body may be a mobile body that moves autonomously based on operational commands. A mobile body may be a vehicle (e.g., a car, an airplane), an unmanned mobile body (e.g., a drone, an autonomous vehicle), or a robot (manned or unmanned). A mobile station also includes devices that do not necessarily move during communication operations. For example, the mobile station could be an IoT (Internet of Things) device such as a sensor.

[0225] (13) In each of the embodiments and the first to second modifications, the term “decision” may encompass a wide variety of actions. “Decision” may include, for example, judging, calculating, computing, processing, deriving, investigating, looking up, searching, inquiry (e.g., searching in a table, database or another data structure), and ascertaining. “Decision” may also include, for example, receiving (e.g., receiving information), transmitting (e.g., sending information), input, output, and accessing (e.g., accessing data in memory), and ascertaining. “Decision” may also include, for example, resolving, selecting, choosing, establishing, and comparing. In other words, "decision" can include considering that some action has been "decided." Furthermore, "decision" can be reinterpreted as "assuming," "expecting," or "considering."

[0226] (14) In each of the embodiments and the first to second modifications, the term “connected,” or any variation thereof, means any direct or indirect connection or coupling between two or more elements, and may include the presence of one or more intermediate elements between two elements that are “connected” or “coupled” to each other. The coupling or connection between elements may be physical, logical, or a combination thereof. For example, “connection” may be read as “access.” As used in the present disclosure, two elements may be considered to be “connected” or “coupled” to each other using at least one of one or more wires, cables, and printed electrical connections, and, in some non-limiting and non-exclusive examples, electromagnetic energy having wavelengths in the radio frequency domain, microwave domain, and optical (both visible and invisible) domain.

[0227] (15) In the first embodiment and each of the first to second modifications, the phrase "based on" does not mean "based solely on" unless otherwise specified. In other words, the phrase "based on" means both "based solely on" and "based at least on".

[0228] (16) Any reference to elements using the designations “first” and “second” as used in this disclosure does not limit the quantity or order of those elements in general. These designations may be used in this disclosure as a convenient way to distinguish between two or more elements. Accordingly, references to the first and second elements do not imply that only two elements may be adopted or that the first element must precede the second element in any way.

[0229] (17) Where “include,” “including,” and variations thereof are used in the Embodiments and the First to Second Modifications, these terms are intended to be inclusive, as are the terms “comprising.” Furthermore, where the terms “or” are used in the Embodiments and the First to Second Modifications, they are not intended to be exclusive OR.

[0230] (18) Where articles are added in the translation, for example, a, an, and the in English, the disclosure may include the fact that the noun following these articles is plural.

[0231] (19) The information, parameters, etc. described in this disclosure may be expressed using absolute values, relative values ​​from a given value, or other corresponding information.

[0232] (20) In this disclosure, the term “A and B are different” may mean “A and B are different from each other.” The term may also mean “A and B are each different from C.” Terms such as “separate” and “combined” may be interpreted in the same way as “different.”

[0233] (21) Each aspect / embodiment described herein may be used individually, in combination, or switched between as needed during implementation. Furthermore, notification of certain information (e.g., notification that "it is X") is not limited to explicit notification, but may also be implicit (e.g., by not providing such notification).

[0234] (22) It will be obvious to those skilled in the art that the present invention is not limited to the embodiments described herein. The present invention can be implemented in modified and altered forms without departing from the spirit and scope of the invention as defined by the claims. Accordingly, the descriptions herein are for illustrative purposes only and are not intended to be restrictive in any way to the present invention. Furthermore, multiple embodiments selected from those illustrated herein may be combined.

[0235] 4. Aspects that can be understood from the above-described forms or modifications The following aspects can be understood from at least one of the above-described forms or modifications.

[0236] 4-1: First Embodiment The question determination device according to the first embodiment includes a category determination unit, a related category determination unit, and a question determination unit. The category determination unit determines at least one category to which at least one answer candidate belongs, based on at least one answer candidate related to an inquiry from a user. The related category determination unit determines at least one related category related to the inquiry from among the at least one category. The question determination unit determines a question relating to the purpose of the inquiry by using the at least one related category.

[0237] According to this embodiment, it becomes possible to provide users with questions that do not include classification symbols such as "Chapter 1, Section 2." Therefore, the possibility of questions becoming redundant questions that include classification symbols such as "Chapter 1, Section 2" can be reduced.

[0238] In addition to inquiries, the system also uses potential answers that the user may not be aware of to determine the appropriate question. Therefore, it can determine an appropriate question regardless of the user's skill level.

[0239] Furthermore, at least one category to which at least one answer candidate belongs succinctly expresses the gist of at least one answer candidate. Therefore, even if an answer candidate that the user is unfamiliar with is used, it is possible to determine a question that is easy for the user to understand.

[0240] Furthermore, even if at least one category includes categories unrelated to the query, the related category determination unit will determine at least one related category that does not include categories unrelated to the query. This reduces the possibility of a question being determined based on categories unrelated to the query. In addition, the related category determination unit can simultaneously eliminate categories with low relevance and identify related categories relevant to the question.

[0241] Furthermore, it eliminates the need to prepare sets of questions and answers in advance. This reduces the workload required to prepare these sets. Additionally, systems that use sets of questions and answers have the problem of standardized answers. In this embodiment, since sets of questions and answers are not required, inquiries can be handled flexibly.

[0242] 4-2: Second Embodiment In the example of the first embodiment (second embodiment), when the at least one related category includes multiple related categories related to the inquiry, the question determined by the question determination unit is a question that prompts the user to select one related category related to the purpose of the inquiry from among the multiple related categories. According to this embodiment, the user only needs to respond by selecting one related category from among the multiple related categories. Therefore, a question that is easy for the user to answer can be used as the question.

[0243] 4-3: Third Embodiment In an example of the first or second embodiment (third embodiment), the question determined by the question determination unit when the at least one related category is one related category related to the inquiry is a question that prompts the user to confirm that the one related category is related to the purpose of the inquiry. According to this embodiment, the user only needs to respond by confirming one related category. Therefore, the burden on the user to respond can be reduced. In addition, it can be prevented from having the user select one related category more than necessary.

[0244] 4-4: Fourth Embodiment In any example of the first to third embodiments (fourth embodiment), the at least one category is generated by the generating AI in response to an instruction from the category determination unit, and the instruction is an instruction to generate the at least one category based on the at least one answer candidate. According to this embodiment, at least one category can be generated by using the generating AI.

[0245] 4-5: Fifth Embodiment In any example of the first to fourth embodiments (fifth embodiment), the related category determination unit determines the at least one related category based on relatedness information indicating the degree of relatedness between the at least one category and the query. According to this embodiment, at least one related category can be determined based on the degree of relatedness between the at least one category and the query.

[0246] 4-6: Sixth Embodiment In the example of the fifth embodiment (sixth embodiment), the relevance information is generated by the generating AI in response to an instruction from the related category determination unit, and the instruction is an instruction to generate the relevance information based on the at least one category and the query. According to this embodiment, relevance information can be generated by using the generating AI.

[0247] 4-7: Seventh Embodiment In any example of the first to sixth embodiments (seventh embodiment), the question determination device further includes a question presentation unit that presents the question to the user using a display device, and a purpose determination unit that determines the purpose of the inquiry based on the response to the question, the question, and the inquiry. According to this embodiment, for example, even if the purpose of the inquiry cannot be determined by the inquiry alone, the purpose of the inquiry can be determined based on the inquiry, the question, and the response.

[0248] 4-8: Eighth Embodiment Example of the Seventh Embodiment (Eighth Embodiment) The objective determination unit determines the objective of the inquiry based on the sentence generated by the generating AI in response to an instruction to generate a sentence based on the response, the question, and the inquiry. According to this embodiment, information that is missing from the inquiry can be included in the sentence generated by the generating AI based on the dialogue of inquiry, question, and response. Therefore, the accuracy of objective determination can be improved.

[0249] 4-9: The ninth embodiment In any example of the first to eighth embodiments (the ninth embodiment), the question determination device further includes an answer candidate determination unit that determines at least one answer candidate from among a plurality of answer candidates based on the query. According to this embodiment, at least one answer candidate can be determined from among a plurality of answer candidates based on the query.

[0250] 4-10: Tenth Embodiment The question determination method according to the tenth embodiment includes: determining at least one category to which at least one candidate answer belongs, based on at least one candidate answer related to an inquiry from a user; determining at least one related category related to the inquiry from among the at least one category; and determining a question relating to the purpose of the inquiry by using the at least one related category. According to this embodiment, the same effects as the first embodiment can be achieved.

[0251] RS...Response system, 1...Summary unit, 2...Answer candidate determination unit, 3...Category determination unit, 4...Related category determination unit, 5...Question determination unit, 6...Relevance determination unit, 7...Question judgment unit, 8...Purpose inquiry determination unit, 9...Purpose confirmation determination unit, 10...Terminal, 11...Input device, 12...Display device, 13...Communication device, 14...Storage device, 15...Processing device, 151...Operation control unit, 16...Bus, 20...Server, 21...Communication device, 22...Storage device, 22a...Answer candidate information, 23...Processing device, 231...Inquiry reception unit, 232...Purpose determination unit, 232a...Purpose classification unit, 232b...Purpose classification determination unit, 233...Question management unit, 234...Question presentation unit, 235...Response reception unit, 236...Dialogue integration unit, 237...Purpose output unit.

Claims

1. A question determination device comprising: a category determination unit that determines at least one category to which at least one answer candidate belongs based on at least one answer candidate related to a user inquiry; a related category determination unit that determines at least one related category related to the inquiry from among the at least one category; and a question determination unit that determines a question relating to the purpose of the inquiry by using the at least one related category.

2. The question determination device according to claim 1, wherein the question determined by the question determination unit when the at least one related category includes a plurality of related categories related to the inquiry is a question that prompts the user to select one related category related to the purpose of the inquiry from among the plurality of related categories.

3. The question determination device according to claim 1, wherein the question determined by the question determination unit when the at least one related category is one related category related to the inquiry is a question that prompts the user to confirm that the one related category is related to the purpose of the inquiry.

4. The question determination device according to claim 1, wherein the at least one category is generated by a generating AI in response to an instruction from the category determination unit, and the instruction is an instruction to generate the at least one category based on the at least one answer candidate.

5. The question determination device according to claim 1, wherein the related category determination unit determines the at least one related category based on relatedness information indicating the degree of relatedness between the at least one category and the query.

6. The question determination device according to claim 5, wherein the relevance information is generated by a generating AI in response to an instruction from the relevance category determination unit, and the instruction is an instruction to generate the relevance information based on the at least one category and the inquiry.

7. The question determination device according to claim 1, further comprising: a question presentation unit that presents the question to the user using a display device; and a purpose determination unit that determines the purpose of the inquiry based on the response to the question, the question, and the inquiry.

8. The question determination device according to claim 7, wherein the purpose determination unit determines the purpose of the inquiry based on a sentence generated by the generating AI in response to an instruction to generate a sentence based on the response, the question, and the inquiry.

9. The question determination device according to claim 1, further comprising a question candidate determination unit that determines at least one question candidate from among multiple question candidates based on the question.

10. A question determination method comprising: determining at least one category to which at least one candidate answer belongs, based on at least one candidate answer related to a user inquiry; determining at least one related category related to the inquiry from among the at least one category; and determining a question relating to the purpose of the inquiry by using the at least one related category.

Images