Kedar Bellare, Anish Das Sarma, Atish Das Sarma, Navneet Loiwal, Vaibhav Mehta, Ganesh Ramakrishnan, Pushpak Bhattacharyya

Department of Computer Science and Engg. , Indian Institute of Technology, Bombay , Powai, Mumbai, India , _kedarb,anish,atish,navneet,vaibhav,hare,pb_@cse.iitb.ac.in

This paper presents a WordNet based approach to text summarization. The document to be summarized is used to extract a “relevant” sub-graph from the WordNet graph. Weights are assigned to each node of this sub-graph using a strategy similar to the Google Pageranking algorithm. These weights capture the relevance of the respective synsets with respect to the whole document. A matrix in which each row repesents a sentence and each column a node of the sub-graph (i.e., a synset) is created. Principal Component Analysis is performed on this matrix to help extract the sentences for the summary. Our approach is generic unlike most previous approaches which address specific genres of documents like news articles and biographies. Testing our system on the standard DUC2002 extracts shows that our results are promising and comparable to existing summarizers.