Lexical Retrieval

Theoretical Foundations
Lexical Retrieval
Lexical Retrieval is a search method based on exact word and phrase matching (BM25, TF-IDF) — fast and cheap but can't understand synonyms.

Lexical Retrieval is a search method based on exact word and phrase matching, using BM25 and TF-IDF algorithms. It's fast and computationally cheap, but can't understand synonyms or intent (the Vocabulary Mismatch problem).

In Google's hybrid system, lexical retrieval serves as the 'preliminary exam', selecting the top ~100 candidates from billions of pages. For SEO, this means using exact keywords is still important — it's the first filter you must pass. Lexical retrieval relies on the Inverted Index, where each word is mapped to a list of documents.

In practice, don't abandon exact keyword phrases in favor of 'writing naturally' — without them you won't pass the initial BM25 filter and won't even reach the top 100 pool from which reranking selects the top 10. Lexical retrieval and semantic retrieval work in parallel within Google's hybrid system.

Source: AI Semantic SEO Expert, Robert Niechciał (sensai.io)