Neural Matching (2018)

Neural Matching (2018) is Google's system that connects queries with documents at the concept level rather than word level. It understands queries even when pages don't contain identical phrases — linking 'why does my monitor flicker' with a page about 'screen refresh rate'. Neural Matching solves the Vocabulary Mismatch problem where users and authors use different words for the same concept.

Neural Matching bridges RankBrain (2015) and BERT (2019) in Google's evolution toward semantic understanding. Neural Matching's success highlights the importance of the lexical expansion tools (which generate synonyms, hyponyms, and meronyms) — they expand the range of queries that can match relevant content.

This system rewards content that covers concepts from multiple angles. Content benefits from varied terminology (swimming pool, pool, natatorium) as Neural Matching recognizes these conceptual connections and broadens query coverage.