A framework for effective retrieval

Publisher Website
Google Scholar
Abstract
The aim of an effective retrieval system is to yield high recall and precision (retrieval effectiveness). The nonbinary independence model, which takes into consideration the number of occurrences of terms in documents, is introduced. It is shown to be optimal under the assumption that terms are independent. It is verified by experiments to yield significant improvement over the binary independence model. The nonbinary model is extended to normalized vectors and is applicable to more general queries. Various ways to alleviate the consequences of the term independence assumption are discussed. Estimation of parameters required for the nonbinary independence model is provided, taking into consideration that a term may have different meanings.

This publication has 32 references indexed in Scilit: