A postings list is the inverted-index entry for a term: it records the document identifiers, and possibly positions, where that term occurs. Together, these lists let a search engine retrieve matching documents without scanning the entire collection. section 2.1.3 section 13.2.1
A lexical search pipeline uses postings lists to:
- Normalize and tokenize document text, then store references from tokens to their postings. section 13.2.1
- Tokenize and normalize a query, look up matching postings, and retrieve document hits. section 13.2.1
- Support term-sequence and Boolean matching through document-set operations; positional data enables positional matching. section 2.1.3
- Provide inputs for BM25 scoring, which ranks documents by query-term and document-feature matches. section 13.2.1
A small postings index
For documents such as “red shoes” and “blue shoes,” normalization and tokenization can produce entries like shoes → [(document 1, position 2), (document 2, position 2)]. A query for “shoes” follows that entry to retrieve both documents, after which BM25 can rank the hits. The structure is fast and useful for lexical matching, but it primarily reflects the presence and count of query-term strings rather than their meaning. section 13.2.1
