A Comparison of Two Methods For Soft Boolean Operator Interpretation In Information Retrieval

Experimental studies aimed at evaluating one family of Boolean logic methods are described, in particular, a parameterized fuzzy-logic approach is contrasted with the norm interpretation and regression analysis supports expected results of parameter settings and gives further insight into why the P-norm scheme is superior.

Abstract

Information retrieval systems generally are given Boolean logic queries by users or search intermediaries, in order that an efficient and effective search for relevant documents can be automatically carried out. Previous work with an extended interpretation of Boolean queries has shown that a dramatic improvement in search effectiveness results. Using the P-norm to compute distance from the ideal points in a multi-dimensional space of truth values leads to best results .1 values are on the order of 1 to 4. Other schemes besides the P-norm approach have been proposed in recent years. This paper describes experimental studies aimed at evaluating one family of such methods. In particular, a parameterized fuzzy-logic approach is contrasted with the norm interpretation. Regression analysis supports expected results of parameter settings and gives further insight into why the P-norm scheme is superior.