Bilkent University
Department of Computer Engineering


Cluster Labeling Using External Resources with Semantic Relatedness Filtering


Melih Baydar
MS Student
Computer Engineering Department
Bilkent University

Web search results become more accessible if they are organized by clustering. Cluster labeling, i.e. accurate description of clusters, is an important task in search result presentation. Most labeling methods use feature selection methods for important term extraction. However, the best explanatory labels may be unavailable in the cluster members. Use of the selected terms as queries to find better explanatory labels by using external resources is a possible way of improvement. In this study, we employ external resources and introduce a novel approach that uses semantically related terms to filter candidate labels to obtain more accurate content descriptions. We use several test collections to judge if improvements made by our approach are statistically better than some popular baselines.


DATE: 07 March, 2016, Monday @ 15:40