Sphere of Influence Model in Information Retrieval

Abstract

We present a new method based on fuzzy proximity for scoring and ranking relevant documents according to the user's information need. Our study takes place in the information retrieval domain and uses the terms localization in documents, we assume that the more the occurrences of the query terms are found close to each other in a document the more this document must be in the top of the retrieved document list. After a brief description of some traditional models and the few approaches which use proximity between the occurrences of terms, we present our "sphere of influence" model which scores documents according to the fuzzy proximity between the query terms in them. First, we detail this model and then we demonstrate that our model includes the traditional ones.