When a search returns results that are ordered, the problem of relevance ranking arises in trying to determine which items should be given preference over others.
Inverse document frequency is where terms which occur in relatively fewer documents in the whole collection are given greater weight
Assumes that documents containing the searched word/phrase more frequent has a higher relevancy compared to other documents.
For example, lets that the case where you do a search for ‘contract’ and it brings up two documents. Say document A the keyword ‘contract’ appearing 50 times, while ‘contract’ only appears once in document B. Which document do you think is more relevant to the topic ‘contract’?
NOTE: Under normal circumstances, the above rule should usually stand, however, there’s always one or two exceptions. Say for the above example, if document B only has the word ‘contract’ appearing once and refers to ‘contract’ as ‘agreement’ in the rest of its text, then we would double think before accepting the search engine’s relevancy ranking.
The following article byClarke et al details research on relevance ranking for one to three search terms.