UNH Search FAQ:
Relevance scoring.
author and contact: cwis.admin@unh.edu
updated 08-DEC-1998

Infoseek computes a relevance score between 0% and 100% for each document that matches a search. Our standard Infoseek search form is set to return results sorted by relevance score, with the score shown in the right margin both as a percentage and as a graphical bar.

As an alternative, Infoseek allows sorting results by date, while still showing the relevance score. Some Web authors may implement their own Infoseek search forms with that alternative. You can see an example on our customized search forms demonstration page (in the compact search example).

The scoring algorithm is complex and there are more details and discussion. Document authors can improve the scoring of their documents by appropriate use of META and TITLE elements in the HEAD section of their HTML documents. The Infoseek indexer weights these sections relative to the remainder of the body text of the document. These elements and their weights influence the score of each document and what is put in the document summary. The bottom line is: use an accurate TITLE element and use accurate META keyword and description elements whenever possible. See the discussion of META elements and also the disclaimer on abuse of keyterms.

For the curious, the UNH Intranet collection of documents is using Infoseek's default weights for special categories of text relative to the body text of a document: