Implementation of an Information Retrieval System (ANIRS) with Ranking and Browsing Capabilities

Abstract

This report describes an implementation of a cluster based information retrieval system with statistical ranking facilities, ANIRS. ANIRS uses the vector space model to represent the document database. In this model, the database is defined by a document by term, D, matrix. In this matrix, each row represents the terms in a single document and each column represents the documents that contain a single term. In ANIRS, two matching methodologies are allowed: a full database search and a cluster based search. The system uses a natural language query interface. It incorporates suffix stripping for term conglomeration. Two methods of query refinement are used: relevance feedback and document seed searching. Cluster browsing, the ability to look at all the documents in a single cluster, is also implemented.