Multi-Field & Boosts

Documents have STRUCTURE: title, body, tags, author, etc. Different fields have different importance.

Indexing strategies:

Field-keyed inverted index: separate posting lists per field.

title:cat -> [doc1, doc7]
body:cat  -> [doc1, doc2, doc7, doc15]
tags:cat  -> [doc1]

Single-field with boosts at query time:

score = bm25(title, q) * 3.0  +  bm25(body, q) * 1.0  +  bm25(tags, q) * 2.0

The 3.0 / 1.0 / 2.0 are field boosts — title matches count more than body matches.

Other common scoring tweaks:

• Exact title match: very strong boost when query == title
• Recency boost: newer docs score higher (decay function on doc age)
• Popularity / clickthrough boost: docs users actually clicked rank higher
• Authority (PageRank): linked-to docs rank higher
• User personalization: re-rank based on user's history

These signals combined are how Google's ranking works — hundreds of factors blended into one score. Open-source search engines (Elasticsearch, Solr) expose mechanisms for all this; you tune for your use case.