So far very little has been said about the actual process by which the required information is located. In the case of document retrieval the information is the subset of documents which are deemed to be relevant to the query. In Chapter 4, occasional reference was made to search efficiency, and the appropriateness of a file structure for searching. The kind of search that is of interest, is not the usual kind where the result of the search is clear cut, either yes, the item is present, or no, the item is absent. Good discussions of these may be found in Knuth[1] and Salton[2]. They are of considerable importance when dictionaries need to be set-up or consulted during text processing. However, we are more interested in search strategies in which the documents retrieved may be more or less relevant to the request.

All search strategies are based on comparison between the query and the stored documents. Sometimes this comparison is only achieved indirectly when the query is compared with clusters (or more precisely with the profiles representing the clusters).

The distinctions made between different kinds of search strategies can sometimes be understood by looking at the query language, that is the language in which the information need is expressed. The nature of the query language often dictates the nature of the search strategy. For example, a query language which allows search statements to be expressed in terms of logical combinations of keywords normally dictates a Boolean search. This is a search which achieves its results by logical (rather than numerical) comparisons of the query with the documents. However, I shall not examine query languages but instead capture the differences by talking about the search mechanisms.

Boolean search

A Boolean search strategy retrieves those documents which are 'true'