About Possibilistic Queries and Their Evaluation

In this paper, the issue of querying databases that may contain ill-known values represented by possibility distributions is considered. The queries dealt with, called possibilistic queries, are of the form: "to what extent is it possible that tuple t belongs to the result of query Q," where Q denotes a usual relational query. An evaluation method suited to these queries is proposed that does not entail making explicit the different possible worlds of the database. Therefore, it is possible to avoid the combinatorial growth that would be induced by a "naive" evaluation method. The validity of this strategy is proven by showing that it corresponds to a valid double rewriting of the query Q. The conditions for the method to be valid are i) that Q must not include any difference operation and ii) that several copies of a same relation must not be involved.