Bilkent University
Department of Computer Engineering


Cluster Labeling by External Sources Found by Queries and Data Fusion


Gökçe Ayduğan
MS Student
Computer Engineering Department
Bilkent University

Web search engine outputs can be overwhelming for users. Effective techniques are needed for efficient access of search results. One popular method is clustering. Clusters must be assigned labels to define their contents. Current cluster labeling methods apply statistical techniques for feature selection and “important” terms are used as cluster labels. However, these methods fail due to two reasons: i) The suggested terms, may conflict with each other; ii) A good label may not occur directly in the original text. In this study we address these problems and use external sources for cluster label generation. In this presentation, we focus on the query which is used for external resource identification. We use data fusion techniques to merge several possibilities. Statistical tests are used for comparisons of the suggested methods.


DATE: 29 February, 2016, Monday @ 16:10