| Hi-index | 0.00 |
Graph-based models and ranking algorithms have been drawn considerable attentions from the document summarization community in the recent years. However, in regard to query-oriented summarization, the influence of the query has been limited to the sentence nodes in the previous graph models. We argue that other than the sentence nodes the sentence-sentence edges should also be measured in accordance with the given query. In this paper, we develop a query-sensitive similarity measure that incorporates the query influence into the evaluation of sentence-sentence edges for graph-based query-oriented summarization. Furthermore, in order to cope with the multi-document summarization task, we explicitly distinguish the inter-document sentence relations from the intra-document sentence relations and emphasize the influence of global information from the document set on local sentence evaluation. Experimental results on DUC 2005 dataset are quite promising and motivate us to further investigate query-sensitive similarity measures.