End-to-end product knowledge graph pipeline on 75K real Amazon ESCI listings: BERT NER (F1=0.90, 3 entity types), semi-supervised expansion, bi-encoder deduplication, knowledge distillation (1.99x speedup), NetworkX/Neo4j graph.

Build a product knowledge graph from Amazon ESCI data using BERT entity extraction, model distillation, and graph construction tools.

Two-stage retrieve-and-rank neural product search on Amazon ESCI: a dense bi-encoder retriever with hard-negative mining + a DeBERTa cross-encoder reranker over Exact/Substitute/Complement/Irrelevant labels. NDCG@10 0.71 (+16% vs BM25), 0.74 micro-F1.