Best Match Retrieval

The central problem in information retrieval is matching a user's query to records in the database. There is an ideal, unattainable search on a perfect system which will find all the items which match the user's search - and no others. In practice, this does not happen. There are cases where the search results in items that the user does not want and cases where the search results in the system not retrieving items relevant to the user's search requirements.

In dealing with this problem, two kinds of retrieval can be contrasted.

Boolean retrieval. This requires a user to construct boolean queries in order to retrieve the items they want. Boolean systems are exact match systems since only records which match the query exactly are retrieved asa set. Every search results in a set of records but this may be an empty set).

Probabilistic retrieval. This requires a user to type a query as an unstructured list of words. Probabilistic systems are best match systems since records are retrieved in a ranked order in terms of how well they match the user's query and how relevant they are likely to be.