Information Extraction and Knowledge Graph Labeling Contributor
Participated in the construction of an intelligent troubleshooting system by extracting structured knowledge from multimodal PDF documents. Used a lightweight PP-UIE-base information extraction model to build a 'fault-cause-action' knowledge graph within the equipment maintenance domain. Labeled entities and relationships from technical documentation to enable accurate query and retrieval workflows. • Parsed and segmented high-precision PDF layouts into semantic chunks • Labeled entities relevant to faults, causes, and actions for a knowledge graph • Utilized Neo4j for graph construction and bge-m3 for metadata embedding • Integrated labeling output into hybrid search and self-check mechanisms