Graded-Inclusion-Based Information Retrieval Systems

Abstract

Abstract. This paper investigates the use of fuzzy logic mechanisms coming from the database community, namely graded inclusions, to model the information retrieval process. In this framework, documents and queries are represented by fuzzy sets, which are paired with operations like fuzzy implications and T-norms. Through different experiments, it is shown that only some among the wide range of fuzzy operations are relevant for information retrieval. When appropriate settings are chosen, it is possible to mimic classical systems, thus yielding results rivaling those of state-of-the-art systems. These positive results validate the proposed approach, while negative ones give some insights on the properties needed by such a model. Moreover, this paper shows the added-value of this graded inclusion-based model, which gives new and theoretically grounded ways for a user to easily weight his query terms, to include negative information in his queries, or to expand them with related terms.

Different interpretations of fuzzy gradual-inclusion-based IR models

“…This point has been experimentally verified: experiments in [3] have shown that R-implications can give good results only when weights are chosen such that the thresholds are never reached.…”

“…An IR system, founded on this fuzzy model has been implemented and tested on different standard collections of documents [3,14]. It was parametrized using a terms weighting scheme adapted from the one in BM25 (which is one of the best), normalized to fit properties of membership degrees.…”

“…However, these works were theoretical studies only, and this kind of IR fuzzy model has never been experimentally tested and validated until the work reported in [3].…”

“…R-implication and S-implications), while the conjunction min seems to be set by theory. However, it has been shown in [3] that the efficiency of the fuzzy model closely depends on the chosen operators. Moreover, the semantics of the query terms weights µ q (t) depends on the chosen family of implications (R-or S-implication).…”

“…However, recent studies in the field of flexible DB querying lead to new querying mechanisms which are more suited to IR. Moreover, from the work by Bosc et al [1,2] on the division of fuzzy relations, new fuzzy IR models based on gradual inclusion have been proposed and experimentally validated [3,4]. The considered gradual inclusions are founded either on an implication or on a cardinality measure.…”