Automatic Clustering of Papers in Thematic Fields and Working Sessions

This article proposes an approach for semi-automatic paper clustering that suggests: which papers to be grouped together in common working sessions, the number of sessions (thematic fields), and the inter-session and intra-session organization, i.e. the order of sessions and the order of papers within a session. In contrast to the existing solutions this approach does not try to group papers in a predefined number of equally-sized sessions, but follows the natural data fragmentation to achieve higher accuracy by producing integral and consistent clusters, independent on the number of target sessions and their size. Experimental results show that the proposed solution is feasible and provides very good initial grouping of papers that greatly facilitates the preparation of the conference program.