Modeling the retrieval process for an information retrieval system using an ordinal fuzzy linguistic approach

A linguistic model for an Information Retrieval System (IRS) defined using an ordinal fuzzy linguistic approach is proposed. The ordinal fuzzy linguistic approach is presented, and its use for modeling the imprecision and subjectivity that appear in the user-IRS interaction is studied. The user queries and IRS responses are modeled linguistically using the concept of fuzzy linguistic variables. The system accepts Boolean queries whose terms can be weighted simultaneously by means of ordinal linguistic values according to three possible semantics: a symmetrical threshold semantic, a quantitative semantic, and an importance semantic. The first one identifies a new threshold semantic used to express qualitative restrictions on the documents retrieved for a given term. It is monotone increasing in index term weight for the threshold values that are on the right of the mid-value, and decreasing for the threshold values that are on the left of the mid-value. The second one is a new semantic proposal introduced to express quantitative restrictions on the documents retrieved for a term, i.e., restrictions on the number of documents that must be retrieved containing that term. The last one is the usual semantic of relative importance that has an effect when the term is in a Boolean expression. A bottom-up evaluation mechanism of queries is presented that coherently integrates the use of the three semantics and satisfies the separability property. The advantage of this IRS with respect to others is that users can express linguistically different semantic restrictions on the desired documents simultaneously, incorporating more flexibility in the user–IRS interaction.