Answer generation method and answer generation program

Abstract

This invention provides a response generation method that utilizes LLM or RAG while preventing the leakage of confidential information to external parties. [Solution] The method includes determining whether an input question is related to confidential information, obtaining information about the storage location of the confidential information as an answer to the input question if it is determined that the input question is related to confidential information, obtaining one or more document data related to the question from a database containing multiple document data if it is determined that the question is not related to confidential information, generating a prompt to be input to the conversational AI based on the question and the one or more document data, inputting the prompt to the conversational AI, obtaining the output of the conversational AI as an answer to the input question, and displaying the obtained answer.

Description

Technical Field

[0001] The present disclosure relates to a response generation method and a response generation program.

Background Art

[0002] Patent Document 1 discloses a technique for generating a chatbot specialized in a given technical field by using large language models (LLMs) based on the given technical field.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] By further utilizing Retrieval Augmented Generation (RAG) using LLMs generated by machine learning in the technique disclosed in Patent Document 1, it is conceivable to provide an interactive AI with non-public information accumulated in a database such as an in-house system as a data source. However, providing non-public information to an interactive AI is not preferable from the perspective of security.

[0005] An object of the present disclosure is to provide a response generation method that utilizes RAG while preventing the leakage of non-public information to the outside.

Means for Solving the Problems

[0006] The method according to an embodiment of the present disclosure is to determine whether the input question corresponds to a question regarding non-public information, If it is determined that the entered question is related to confidential information, the information of the storage location of the confidential information will be obtained as the answer to the entered question. If it is determined that the above question does not constitute a question concerning confidential information, From a database containing multiple document data, retrieve one or more document data related to the above question. A prompt to be input to the conversational AI is generated based on the aforementioned question and the one or more document data, Enter the prompt into the conversational AI, The output of the aforementioned conversational AI is obtained as the answer to the input question, Display the aforementioned response obtained, Includes. [Effects of the Invention]

[0007] According to this disclosure, it is possible to generate responses using RAG while preventing the leakage of confidential information to external parties. [Brief explanation of the drawing]

[0008] [Figure 1] This figure shows the configuration of a response generation system according to one embodiment of the present disclosure. [Figure 2] This flowchart shows an example of the operation of a response generation system according to one embodiment of this disclosure. [Modes for carrying out the invention]

[0009] Hereinafter, an embodiment of the present disclosure will be described with reference to the drawings. In each drawing, parts having the same configuration or function are denoted by the same reference numerals. In the description of this embodiment, redundant descriptions of the same parts may be omitted or simplified as appropriate.

[0010] Referring to Figure 1, the configuration of the response generation system 1 according to one embodiment of this disclosure will be described.

[0011] The response generation system 1 comprises an information processing device 10, a first server 20, and a second server 30.

[0012] The information processing device 10 comprises a control unit 11, a storage unit 12, a communication unit 13, an input unit 14, and an output unit 15. The information processing device 10 is a general-purpose computer such as a workstation, personal computer, smartphone, or tablet. The information processing device 10 receives a question from a user, determines whether the question concerns confidential information, receives appropriate information from the first server 20 and the second server 30, and generates an answer to the question.

[0013] The control unit 11 includes one or more processors, one or more dedicated circuits, or a combination thereof, and executes predetermined processing. The processor is a general-purpose processor such as a CPU (Central Processing Unit) or GPU (Graphics Processing Unit), or a dedicated processor specialized for a specific processing. The dedicated circuit is, for example, an FPGA (Field-Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit). The control unit 11 executes various processing related to the operation of the information processing device 10 and controls each part of the information processing device 10.

[0014] The storage unit 12 includes at least one semiconductor memory, at least one magnetic memory, at least one optical memory, or at least two combinations thereof. The storage unit 12 stores data used for the operation of the information processing device 10 and data obtained by the operation of the information processing device 10. In one embodiment of this disclosure, the storage unit 12 stores a record of the interaction between the information processing device 10 and the second server 30. The storage unit 12 may also store a text generation AI that generates text without communicating with the outside world.

[0015] The communication unit 13 includes at least one communication interface. The communication unit 13 receives data used for the operation of the information processing device 10 and transmits data obtained by the operation of the information processing device 10. In one embodiment of the present disclosure, the communication unit 13 communicates with the first server 20 and the second server 30 via a network.

[0016] The input unit 14 includes one or more input interfaces. The input interface is, for example, a physical key, a capacitive key, a pointing device, a touch screen provided integrally with a display, or a microphone that receives voice input. The input unit 14 receives an operation for inputting information used for the operation of the information processing device 10.

[0017] The output unit 15 includes one or more output interfaces. The output interface is, for example, an external or built-in display that outputs information as an image or video, or a connection interface to an external output device. The display is, for example, an LCD (Liquid Crystal Display) or an organic EL (Electro Luminescence) display. The output unit 15 outputs information obtained by the operation of the information processing device 10.

[0018] The first server 20 is a server device installed in, for example, a data center or the like. The first server 20 may be, for example, an in-house system that is not externally connected. The first server 20 includes a database that aggregates in-house product information and in-house information including non-public information. The first server 20 transmits information or information on the storage location of the information to the information processing device 10 in response to a request from the information processing device 10.

[0019] The second server 30 is a server device installed in, for example, a data center or the like. The second server 30 includes a generative AI. The second server 30 inputs the information received from the information processing device 10 and the like into the generative AI, generates a response sentence, and transmits it to the information processing device 10. The generative AI includes, for example, any dialogical AI that uses an LLM.

[0020] Hereinafter, a response generation method according to an embodiment of the present disclosure will be described.

[0021] When the control unit 11 of the information processing apparatus 10 receives a question from the user via the input unit 14, it determines whether the question is related to non-public information. The determination method is not particularly limited. For example, it may be a method of checking whether the question sentence contains keywords related to non-public information.

[0022] When the question is related to non-public information, the control unit 11 of the information processing apparatus 10 searches for and extracts information related to the question from the database provided in the first server 20. The control unit 11 acquires information on the storage location of the extracted information from the first server 20. The information on the storage location is, for example, a file path, a link, a URL, or the like. Thereafter, the control unit 11 generates a sentence indicating that the answer to the question is the information on the storage location, and outputs it to the output unit 15 as the answer to the question.

[0023] When the question is not related to non-public information, the response generation system 1 generates an answer to the input question by utilizing RAG. Specifically, the control unit 11 of the information processing apparatus 10 searches for information related to the question from the database provided in the first server 20 and extracts one or more pieces of information. The extracted information is, for example, document data, but is not limited thereto, and may be, for example, image data including text data. Thereafter, the control unit 11 creates a prompt that the generation AI provided in the second server 30 can handle based on the question, and transmits it to the second server 30 together with the information acquired from the first server 20. The second server 30 generates an answer using the generation AI based on the prompt and information received from the information processing apparatus 10. The control unit 11 of the information processing apparatus 10 outputs the answer received from the second server 30 to the output unit 15.

[0024] Hereinafter, referring to FIG. 2, a response generation method using the response generation system 1 according to an embodiment of the present disclosure will be described. FIG. 2 is a flowchart showing an operation example of the response generation system 1 according to an embodiment of the present disclosure.

[0025] In step S1, the control unit 11 of the information processing device 10 receives a question from the user via the input unit 14. The control unit 11 may accept the question in a normal text format or in a prompt format.

[0026] In step S2, the control unit 11 of the information processing device 10 determines whether the question received in step S1 is related to confidential information. For example, if the received question contains keywords that have been set in advance as being related to confidential information, it may be determined to be a question related to confidential information. If the control unit 11 determines that the question is related to confidential information (step S2: Yes), the process proceeds to step S3. If the control unit 11 determines that the question is not related to confidential information (step S2: No), the process proceeds to step S6.

[0027] In step S3, the control unit 11 of the information processing device 10 searches and extracts information related to the question received in step S1 from the database provided by the first server 20.

[0028] In step S4, the control unit 11 of the information processing device 10 obtains information about the storage location of the information extracted in step S3 from the first server 20.

[0029] In step S5, the control unit 11 of the information processing device 10 generates a statement indicating that the answer to the question received in step S1 is the storage location information obtained in step S4, outputs this statement as the answer to the output unit 15, and terminates the process. The statement generated as the answer may be, for example, "This is confidential information, so please check the <storage location information>."

[0030] In step S6, the control unit 11 of the information processing device 10 searches and extracts information related to the question received in step S1 from the database provided by the first server 20.

[0031] In step S7, the control unit 11 of the information processing device 10 acquires the information extracted in step S3 from the first server 20.

[0032] In step S8, the control unit 11 of the information processing device 10 generates a prompt that can be handled by the generating AI in the second server 30, based on the question received in step S1. The control unit 11 sends the generated prompt and the information acquired in step S6 to the second server 30. The second server 30 receives the prompt and information from the information processing device 10.

[0033] In step S9, the second server 30 inputs the prompt and information received from the information processing device 10 into the generating AI and generates a response. Once the second server 30 has generated a response, it sends it to the information processing device 10.

[0034] If the prompt and information received by the second server 30 in step S8 are insufficient to generate an answer, the second server 30 may query the information processing device 10 for the missing information. When the control unit 11 of the information processing device 10 receives a query from the server device 30, it may generate a prompt that will be the answer to the query based on the question, or obtain the necessary information from the first server, and send it to the second server 30. If the information required to create a prompt that will be the answer to the query is insufficient from the question alone, the control unit 11 of the information processing device 10 may output a message to the output unit 15 requesting supplementary information for the question. When supplementary information for the question is input to the input unit 14, the control unit 11 may generate a prompt or obtain the necessary information from the first server 20 based on the supplementary information, and send it to the second server 30.

[0035] The control unit 11 of the information processing device 10 may complete the question based on past dialogue records with the conversational AI provided by the second server 30. The timing of the completion is not particularly limited, but for example, the control unit 11 of the information processing device 10 may complete the question when it receives an inquiry from the second server 30.

[0036] In step S10, when the control unit 11 of the information processing device 10 receives a response from the second server 30, it outputs the response to the output unit 15 and terminates the process.

[0037] By employing the above method, for questions related to confidential information, answers can be generated solely through information sharing between the information processing device 10 and the first server 20, thereby reducing the risk of confidential information being leaked externally through the use of the generation AI. On the other hand, for questions not related to confidential information, accurate answers can be generated quickly by utilizing RAG.

[0038] This disclosure is not limited to the embodiments described above. For example, multiple blocks described in the block diagram may be combined, or a single block may be divided. Instead of executing multiple steps described in the flowchart in chronological order as described, they may be executed in parallel or in a different order, depending on the processing capacity of the device performing each step, or as necessary. Other modifications are possible without departing from the spirit of this disclosure.

[0039] Furthermore, if the first server is an internal company system, the control unit 11 of the information processing device 10 may search for and extract information that the internal company system stores outside the database. [Explanation of Symbols]

[0040] 1. Answer Generation System 10 Information Processing Devices 11 Control Unit 12 Storage section 13 Communications Department 14 Input section 15 Output section 20 Server 1 30 Second Server

Claims

1. A method for generating an answer that is performed by an information processing device, To determine whether the entered question falls under the category of a question about confidential information, If it is determined that the entered question is related to confidential information, the information of the storage location of the confidential information will be obtained as the answer to the entered question. If it is determined that the above question does not constitute a question concerning confidential information, From a database containing multiple document data, retrieve one or more document data related to the above question. A prompt to be input to the conversational AI is generated based on the aforementioned question and the one or more document data, Enter the prompt into the aforementioned conversational AI, The output of the aforementioned interactive AI is obtained as the answer to the input question, Display the aforementioned response obtained, A method for generating answers, including the method itself.

2. The response generation method according to claim 1, wherein the information of the storage location of the confidential information is aggregated in the database.

3. The method for generating an answer according to claim 1, wherein obtaining information about the storage location of the confidential information as an answer to the input question includes searching for and obtaining information about the storage location of the confidential information from an in-house system to which the information processing device is connected.

4. The recording of the conversation with the aforementioned interactive AI is stored, To supplement the aforementioned questions based on the aforementioned dialogue record, The method for generating an answer according to claim 1, further comprising:

5. It is an answer generation program, In an information processing device, To determine whether the entered question falls under the category of a question about confidential information, If it is determined that the entered question is related to confidential information, the information of the storage location of the confidential information will be obtained as the answer to the entered question. If it is determined that the above question does not constitute a question concerning confidential information, From a database containing multiple document data, retrieve one or more document data related to the above question. A prompt to be input to the conversational AI is generated based on the aforementioned question and the one or more document data, Enter the prompt into the aforementioned conversational AI, The output of the aforementioned interactive AI is obtained as the answer to the input question, Display the aforementioned response obtained, A response generation program that performs processing that includes this process.

Images