Properties of topologies of information retrieval systems

This paper studies topological properties of different topologies that are possible on the space of documents as they are induced by queries in a query space together with a similarity function between queries and documents. The main topologies studied here are the retrieval topology (introduced by Everett and Cater) and the similarity topology (introduced by Egghe and Rousseau). The studied properties are the separation properties T-0, T-1, and T-2 (Hausdorff), proximity and connectedness. Full characterizations are given for the diverse topologies to be T-0, T-1 or T-2. It is shown that the retrieval topology is not necessarily a proximity space, while the similarity topology and the pseudo-metric topology always are proximity spaces. A characterization of connectedness in terms of the Boolean NOT-operator is given, hereby showing the intimate relationship between IR and topology.