Pairwise matching treats each comparison as a one-off. A persistent knowledge graph turns product mentions, manufacturers, model numbers, attributes, and price bins into typed nodes and relations. Matching becomes neighborhood comparison: multi-hop paths (convergent evidence) can beat any single similarity score.

False entity merges don’t just dirty data. They distort inventory, pricing, and forecasts, then every model and report built on top. Learned sparse retrieval improves recall, but it can still treat records like unordered tokens. This post adds token-to-token attention as a structural check so near-duplicates pass and lookalikes fail, with a trail you can audit.