CILKS Final Report : Query Enhancement

What is Query Enhancement?

A query is an expression of a user's information need, intended to identify relevant records from a document database. Many initial queries consist of only one or two terms, but they may require enhancement with additional terms in order to ensure adequate:

recall: a good proportion of relevant documents is found,
precision: not too many irrelevant documents are found.

Traditional queries using the boolean operators (AND, OR, and NOT) may be:

refined: made narrower by specifying additional terms which must be present in retrieved documents.
expanded: made broader by removing some constraints, or specifying additional terms which may be present in retrieved documents.

In probabilistic retrieval, it is hoped that additional terms will both refine and expand a query, by finding extra relevant documents, and also by promoting those which are most relevant (i.e. match the query most closely) to the top of the ranked hit list.

In the CILKS project, all query enhancement via the thesaurus took place before the query was submitted to the Okapi document retrieval system. This allowed tight control of the evaluation experiments. Under normal circumstances, enhancement may take place at any time during a retrieval session, after the user has seen the documents retrieved by his or her initial query.