PROJECT 1:
Construction of a Lithium Battery Knowledge Base Based on LLM
Principal Researchers: 吴盈盈 (Yingying Wu, 25 Fall Mater); 胡开泉 (Kaiquan Hu, 25 Fall Mater)
1. OVERALL SYSTEM ARCHITECTURE
The figure below illustrates the end-to-end architecture of our RAG (Retrieval-Augmented Generation) system. It details the complete workflow from raw document preprocessing and vectorization (`bge-m3`), to multi-stage precision retrieval, and finally to LLM-driven intelligent answer generation.
2. LLM-POWERED DATA MINING PIPELINE
This pipeline demonstrates our multi-agent approach to data mining. By utilizing sequential stages—Material Discovery, Rule Screening, and Unified Extraction—the system efficiently filters out noise and transforms unstructured scientific texts into highly structured JSON data formats.
3. KNOWLEDGE BASE STATISTICS & VISUALIZATIONS
Based on the constructed knowledge base, we conducted deep statistical analysis. The charts highlight key trends, including the frequency of extracted material properties and battery performance metrics, as well as the elemental composition distribution of anode materials.
