Natural Language Database Access Using a Continuously Updated RAG Layer
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems require specialized knowledge to access and analyze information stored in structured databases like RDBs, limiting effective use by users without such expertise.
Innovation Solution
An information processing system that enables natural language interaction with a structured database by using a large language model (LLM) through a retrieval-augmented generation (RAG) database, converting database queries into natural language responses and updating the RAG database with predefined conversion rules.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a large-scale language model is used to enable natural language interaction with databases, then ease of operation is improved, but the system cannot accurately access and analyze information stored in structured databases without specialized knowledge
Solution Approach 1:
The patent introduces a RAG database as an intermediary between the LLM and the structured database. The RAG database stores pre-processed information in an unstructured format that the LLM can easily access, while a conversion unit translates LLM queries into database queries and converts database results back into natural language responses. This intermediary layer enables the LLM to access database information accurately without requiring it to learn the complex data structure of the structured database.
Solution Approach 2:
The patent implements preliminary action by pre-processing and storing database information in the RAG database before actual queries are made. The conversion unit pre-establishes the mapping between natural language queries and database queries, and the RAG database pre-stores information in a format suitable for LLM processing. This preparation allows the system to accurately retrieve information without requiring the LLM to learn or memorize the structured database schema.
2Device complexity
If the LLM directly accesses structured database information, then device complexity is reduced, but the LLM must learn the data structure which limits its ability to handle daily updated information
Solution Approach 1:
The patent segments the information access pathway into distinct components: the structured database, the conversion unit, and the RAG database. The structured database maintains its complex data structure for efficient storage and updates, while the RAG database maintains a simplified unstructured format for easy LLM access. This segmentation allows each component to optimize for its specific function without compromising the other.
Solution Approach 2:
The RAG database serves as a mediator that decouples the LLM from the structured database. The conversion unit acts as another mediator that translates between natural language queries and database queries. This intermediary architecture allows the system to easily adapt to daily database updates without requiring the LLM to relearn the data structure, as the RAG database can be updated independently to reflect new information.
3Measurement precision
If specialized knowledge is required to access database information, then measurement precision is maintained, but accessibility to users without specialized knowledge is reduced
Solution Approach 1:
The RAG database and conversion unit together form an intermediary system that bridges the gap between precise database information and user-friendly natural language queries. The conversion unit ensures that information retrieval remains precise by accurately translating queries and results, while the RAG database presents information in an accessible format that users without specialized knowledge can easily understand and utilize.
Solution Approach 2:
The patent replaces the mechanical system of direct database querying (which requires specialized SQL knowledge) with a natural language-based system. Instead of requiring users to learn and execute complex queries, the system substitutes this with natural language processing, where users can ask questions in plain language and receive accurate responses through the automated conversion and retrieval processes.
Data Source
AI summary
According to one embodiment, an information processing system includes an information processing device and a structured database accessible by a specialized query language. The information processing device receives a natural language query related to information in the structured database and provides a natural language response to the natural language query by supplying the natural language query to a large language model (LLM) accessing a retrieval-augmented generation (RAG) database incorporating information from the structured database. The information processing device is configured to automatically update information in the RAG database by retrieving information from the structured database by executing a predefined query in the specialized query language to access the structured database, converting results from the predefined query into natural language text according to a predefined conversion rule associated with the predefined query, and storing the natural language text in the RAG database.