RankBrain (2015)

RankBrain (2015) is Google's first machine learning system that interprets new, unknown queries based on similar historical searches. RankBrain marked the beginning of Google's shift away from pure lexical matching toward understanding user intent.

It represents a milestone in search engine evolution — from TF-IDF through BM25, up to BERT and MUM. Despite the introduction of newer systems, RankBrain still operates as part of Google's hybrid ranking stack. RankBrain showed that word matching alone isn't enough — the search engine must understand WHAT the user is looking for, not just WHICH WORDS they use.

In practice, RankBrain interprets new queries through analogy to known ones — that's why covering the semantic field with synonyms, hyponyms, and meronyms increases your chances of matching queries that no one has typed before.