A postings list is the collection of records associated with one term in an inverted index. Each record identifies a document containing the term; optional fields include its term frequency in that document and, for a positional index, occurrence positions. The number of distinct document records is the document frequency, which differs from the total number of occurrences.[ref]
Introduction to Information Retrieval by Christopher D. Manning and others. §1.2, inverted-index construction and document frequency; §2.4.2, positional postings
For documents 1
red red boat, 2 blue boat, and 3 red train, the document identifiers in the red postings list are . Its document frequency is two, although red occurs three times. The boat postings list contains . Their intersection is , answering red AND boat.With identifiers in increasing order, compare the current identifiers and advance whichever is smaller; when equal, emit that identifier and advance both. Each advancement consumes a record, so intersecting lists of lengths costs comparisons in this simple merge algorithm. Ordering gives a direct route from stored evidence to a query result without reading every document's text.[ref]
Introduction to Information Retrieval by Christopher D. Manning and others. §1.3, Processing Boolean queries, Figure 1.6 and the linear intersection bound
New to topics? Read the docs here!