TY - GEN
T1 - Query-oriented summarization based on neighborhood graph model
AU - Wei, Furu
AU - He, Yanxiang
AU - Li, Wenjie
AU - Huang, Lei
PY - 2009/11/9
Y1 - 2009/11/9
N2 - In this paper, we investigate how to combine the link-aware and link-free information in sentence ranking for query-oriented summarization. Although the link structure has been emphasized in the existing graph-based summarization models, there is lack of pertinent analysis on how to use the links. By contrasting the text graph with the web graph, we propose to evaluate significance of sentences based on neighborhood graph model. Taking the advantage of the link information provided on the graph, each sentence is evaluated according to its own value as well as the cumulative impacts from its neighbors. For a task like query-oriented summarization, it is critical to explore how to reflect the influence of the query. To better incorporate query information into the model, we further design a query-sensitive similarity measure to estimate the association between a pair of sentences. When evaluated on DUC 2005 dataset, the results of the pro-posed approach are promising.
AB - In this paper, we investigate how to combine the link-aware and link-free information in sentence ranking for query-oriented summarization. Although the link structure has been emphasized in the existing graph-based summarization models, there is lack of pertinent analysis on how to use the links. By contrasting the text graph with the web graph, we propose to evaluate significance of sentences based on neighborhood graph model. Taking the advantage of the link information provided on the graph, each sentence is evaluated according to its own value as well as the cumulative impacts from its neighbors. For a task like query-oriented summarization, it is critical to explore how to reflect the influence of the query. To better incorporate query information into the model, we further design a query-sensitive similarity measure to estimate the association between a pair of sentences. When evaluated on DUC 2005 dataset, the results of the pro-posed approach are promising.
KW - Neighborhood Graph
KW - Query-Oriented Summarization
KW - Query-Sensitive Similarity
UR - http://www.scopus.com/inward/record.url?scp=70350657186&partnerID=8YFLogxK
U2 - 10.1007/978-3-642-00831-3_15
DO - 10.1007/978-3-642-00831-3_15
M3 - Conference article published in proceeding or book
SN - 3642008305
SN - 9783642008306
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 156
EP - 167
BT - Computer Processing of Oriental Languages
T2 - 22nd International Conference on Computer Processing of Oriental Languages, ICCPOL 2009
Y2 - 26 March 2009 through 27 March 2009
ER -