Intelligent Session Server Table Data Retrieval
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current intelligent session technologies based on text-based knowledge and limited table data struggle to accurately respond to diverse user queries in multi-field and multi-value scenarios, especially with complex tables, and require high memory and computing power for training.
Innovation Solution
An intelligent session method and server that utilizes table data retrieval, involving syntax error correction, intention classification, entity extraction, and a query language conversion model to generate accurate SQL statements, reducing coding complexity and adapting to diverse queries by constructing an entity word library and introducing interaction features into the coding layer.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If text-based knowledge and limited table data are used for intelligent session, then the system is easier to implement, but the accuracy and adaptability to diverse user queries deteriorates
Solution Approach 1:
The system segments the knowledge base into text-based knowledge and table-based data, processing them through different pathways. Text-based queries use natural language processing while table-based queries use SQL generation, allowing each type to be handled optimally without compromising overall accuracy
Solution Approach 2:
An intermediary module is introduced that detects query intent and routes queries to appropriate processing pathways. The system uses intention classification to determine whether a query should be handled by text-based knowledge retrieval or table-based SQL generation, improving accuracy while maintaining implementation simplicity
2Adaptability or versatility
If complex coding layers are used in query language conversion model, then the model can handle more scenarios, but the algorithm accuracy and reasoning efficiency deteriorates
Solution Approach 1:
The coding layer is made dynamic and adaptive rather than static and complex. The system dynamically adjusts coding strategies based on query characteristics and table schemas, using lightweight coding that adapts to different scenarios without requiring overly complex predefined structures
Solution Approach 2:
The system changes parameters such as code length, code complexity, and feature selection based on the specific query and table context. Rather than using fixed complex coding layers, the system adjusts coding parameters dynamically to achieve both scenario coverage and high accuracy
3Power
If high memory and computing power are allocated for training, then the model capacity increases, but the system complexity and resource requirements deteriorates
Solution Approach 1:
The system performs preliminary actions by pre-processing and indexing table data before query time. Tables are pre-analyzed and structured in ways that reduce the computational burden during actual queries, allowing lower runtime resource requirements while maintaining high model capacity
Solution Approach 2:
Complex computational tasks are extracted and separated from the main model. The system extracts table schemas, entity relationships, and query patterns into separate modular components that can be processed independently, reducing the computational burden on the core model while maintaining overall system capability
Data Source
Figure 1~2
Figure 3~4
AI summary
Disclosed are an intelligent session method and server based on table data retrieval. The method includes: invoking an entity extraction and parsing method based on an input corpus to obtain an entity extraction and parsing result, performing feature splicing and feature coding according to table entity word information, and invoking a query language conversion model based on coded features to generate an SQL statement; and submitting to a database execution engine based on the SQL statement to obtain an execution result, and returning the execution result through an interface as a reply of the input corpus. Thus, the problems of poor accuracy of algorithms and low reasoning efficiency caused by high complexity of coding layers in the related art are solved.