A Linguistic Multi-level Weighted Query Language to Represent User Information Needs

An ordinal fuzzy linguistic Information Retrieval System (IRS) based on a multi-level weighting scheme to represent the user queries, in a more flexible way, is proposed. The IRS accepts Boolean queries that can be weighted simultaneously by means of ordinal linguistic values in two weighting levels: level of terms and level of connectives. In level of terms, the weights are associated to a threshold semantics, and in the level of connectives they are associated to a control semantics acting as modifiers of the action of the Boolean classical connectives AND and OR in the retrieval process. A new family of parameterized soft computing operators, called S-LOWA operators, is introduced for modelling that control semantics in the action of the connectives AND and OR.