JP-2026075428-A - Information processing device and program

Abstract

[Problem] To provide an information processing device that can obtain question answers based on a specific information source. [Solution] The information processing device of the present invention comprises: information type designation means for receiving the designation of one or more information types; question receiving means for receiving a user's question from a terminal device; search means for searching for document data of the information type designated by the information type designation means based on the user's question and obtaining document data similar to the user's question; prompt generation means for generating a prompt based on the user's question and the document data retrieved by the search means; response generation means for causing a generating AI to generate an answer based on the prompt and generating a response to the question based on the answer; and response transmission means for transmitting the response generated by the response generation means to the terminal device. [Selection Diagram] Figure 9

Inventors

• 渡辺 圭祐
• 坂 拓弥
• 島袋 誠也

Assignees

• 株式会社野村総合研究所

Dates

Publication Date

20260508

Application Date

20241022

Claims (11)

1. Information type specification means for receiving the specification of one or more information types, A question receiving means that receives user questions from a terminal device, A search means that searches for document data of the information type specified by the information type specification means based on the user's question and obtains document data similar to the user's question, A prompt generation means that generates a prompt based on the user's question and the document data retrieved by the search means, A response generation means that causes the generating AI to generate an answer based on the prompt, and generates a response to the question based on the answer, A response transmission means for transmitting the response generated by the response generation means to the terminal device, An information processing device equipped with the following features.
2. The aforementioned information types indicate types of documents with different levels of confidentiality. The information processing apparatus according to claim 1.
3. The aforementioned information type indicates the type of document in a different field. The information processing apparatus according to claim 1.
4. The response generation means generates a response that includes information from the document data on which the generating AI based its answer. The information processing apparatus according to claim 1.
5. The response generation means generates a response that includes a downloadable object from the document data on which the generating AI is based. The information processing apparatus according to claim 4.
6. The response generation means generates a response including the downloadable object according to the user's access rights. The information processing apparatus according to claim 5.
7. The aforementioned document data is divided into chunks, and if one or more tables exist across multiple chunks, those one or more tables are merged into a single chunk. The information processing apparatus according to claim 1.
8. The aforementioned document data is divided into chunks, and if one or more tables are present across multiple chunks, a header is added to each of the tables included in those chunks. The information processing apparatus according to claim 1.
9. The aforementioned prompt includes a specification to convert dynamic links contained in the document data into accessible links, The information processing apparatus according to claim 1.
10. The response generation means generates the response by converting the dynamic link of the response into an accessible link. The information processing apparatus according to claim 1.
11. A program that causes a computer to function as one of the means of the information processing device described in any one of claims 1 to 10.

Description

This invention relates to an information processing device and a program. In recent years, natural language processing has been performed using deep learning models such as LLM (Large-Scale Language Models). One such method is the Retriever-Reader model, which leverages knowledge accumulated within a company. The Retriever-Reader model uses a question-answering task to retrieve the necessary answers from a vast amount of internal documents. Furthermore, a technology called RAG (Retrieval-Augmented Generation), which combines LLM with the retrieval of information outside of LLM, is known. LLM using RAG makes it possible to utilize internal document knowledge across all business areas accumulated within the company. Non-Patent Literature 1 describes a study in which the Retriever and Reader underwent fine-tuning due to numerous variations in terminology within the knowledge database. The study found that fine-tuning the Retriever to increase the similarity between the question and the document containing the answer improved the accuracy rate. Furthermore, the Reader was found to appropriately determine the answer to a question by dividing the document containing the answer into individual sentences using periods and determining whether each sentence is an answer or not. Hasuda et al., "Verification of a Retriever-Reader Model for the Utilization of Technical Knowledge," Proceedings of the 29th Annual Meeting of the Association for Natural Language Processing, Association for Natural Language Processing, March 6, 2023, pp. 2030-2033. A diagram illustrating the configuration of an information processing system according to an embodiment.A block diagram showing an example of the functional configuration of the information processing device according to the embodiment.A block diagram showing an example of the functional configuration of a terminal device according to this embodiment.A diagram illustrating an example of index creation according to this embodiment.A flowchart illustrating the index creation process according to the embodiment.A diagram illustrating the page splitting process according to the embodiment.A diagram showing an example of a display screen according to the embodiment.A diagram showing an example of settings related to response generation according to the embodiment.A diagram showing the processing sequence of the information processing system according to the embodiment.A diagram showing an example of a display screen according to the embodiment.A diagram illustrating an example of index creation according to this embodiment. <First Embodiment> The embodiments will be described in detail below with reference to the attached drawings. Note that the following embodiments do not limit the invention as defined in the claims, and not all combinations of features described in the embodiments are essential to the invention. Two or more of the features described in the embodiments may be combined in any way. Furthermore, identical or similar configurations will be given the same reference numeral, and redundant descriptions will be omitted. <Configuration of the Information Processing System> Referring to Figure 1, the information processing system according to this embodiment will be described. The information processing system 10 includes, for example, an information processing device 101, an index database (DB) 102, an information source 103, a terminal device 104 used by a user 105, and a generation AI server 106. The information processing device 101, the index DB 102, the information source 103, and the terminal device 104 are connected via the same local area network (LAN) 107. The LAN 107 may be a wired LAN, a wireless LAN, or a combination of both. The LAN 107 is connected to an external network 108 (for example, the Internet) through a firewall 109. The LAN 107 is, for example, an internal network of a company or organization. The terminal device 104, the information processing device 101, the index DB 102, and the information source 103 can communicate with each other within the LAN 107 (i.e., without going through the external network 108). Furthermore, the terminal device 104 and the information processing device 101 can each communicate with the AI generation server 106 via the LAN 107, the firewall 109, and the external network 108. The terminal device 104 may also be connected to the external network 108 via a cellular network (for example, a network compliant with the fourth-generation communication standard (4G) or the fifth-generation communication standard (5G)). The information processing device 101 is a server that functions as the front end of the generation AI server 106. A company or organization may install the information processing device 101 on-premises. The installation configuration of the information processing device 101 is not limited to this. The information processing device 101 and the terminal device 104 can communicate via LAN 107. Access to the information