The University of Michigan in Novelty 2004
نویسنده
چکیده
This year we participated in the Novelty track. To find the relevant sentences, we combine sentence salience features that are inherited from text summarization domain with other heuristic features based on topic statements. We propose a novel method to extract the new sentences based on the graph-based ranking of the similarity relation between the sentences. 1. Overview The University of Michigan participated in all four tasks of the TREC 2004 Novelty track. To find the relevant sentences in Tasks 1 and 3, we experimented with more than ten features. The system was trained with all possible subsets of these features on the Novelty 2003 data using different learning algorithms to be explained below. All of the features were integrated into the MEAD1 text summarization system (Radev, Blair-Goldensohn, & Zhang, 2001). The following is a brief description of the features we used in the actual submissions, which gave us the best results on the training data: • Centroid: The centroid score that is a measure of how close is the sentence to the centroid pseudosentence of the entire cluster. This is a measure of sentence salience which is proven to be successful in multi-document summarization domain (Radev, Jing, & Budzikowska, 2000). • LexRank: The LexRank score (Erkan & Radev, 2004) is a measure of sentence salience based on the eigenvector centrality of the graph-based representation of the sentences in a cluster. We will give a brief explanation of how to compute LexRank in Section 2 and Section 3. • Length: The number of words in the sentence. • QueryTitleCosine: The cosine similarity between the “title” field of the topic statement and the sentence weighted by the word idf’s. Formally, the cosine between two sentences x and y is defined by idf-modified-cosine(x, y) = ∑ w∈x,y tfw,xtfw,y(idfw)2
منابع مشابه
Probing the sustainability of the N=82 and Z=50 shell closures for neutron-rich nuclides:
Decay of Rh75 to levels of Pd74 W. B. Walters, B. E. Tomlin, P. F. Mantica, B. A. Brown, J. Rikovska Stone, A. D. Davies, A. Estrade, P. T. Hosmer, N. Hoteling, S. N. Liddick, T. J. Mertzimekis, F. Montes, A. C. Morton, W. F. Mueller, M. Ouellette, E. Pellegrini, P. Santi, D. Seweryniak, H. Schatz, J. Shergur, and A. Stolz Department of Chemistry and Biochemistry, University of Maryland, Colleg...
متن کاملClinical Report Microcephaly, Jejunal Atresia, Aberrant Right Bronchus, Ocular Anomalies, and XY Sex Reversal
Catherine E. Keegan, Eric Vilain, Mansoor Mohammed, Jessica Lehoczky, William B. Dobyns, Steven M. Archer, and Jeffrey W. Innis* Department of Pediatrics, University of Michigan Medical School, Ann Arbor, Michigan Department of Human Genetics, University of Michigan Medical School, Ann Arbor, Michigan Department of Ophthalmology, University of Michigan Medical School, Ann Arbor, Michigan Depart...
متن کاملEulerian Geometrical Optics and Fast Huygens Sweeping Methods for Three-Dimensional Time-Harmonic High-Frequency Maxwell's Equations in Inhomogeneous Media
In some applications, it is reasonable to assume that geodesics (rays) have a consistent orientation so that Maxwell’s equations may be viewed as an evolution equation in one of the spatial directions. With such applications in mind, we propose a new Eulerian geometrical-optics method, dubbed the fast Huygens sweeping method, for computing Green’s functions of Maxwell’s equations in inhomogeneo...
متن کاملDevelopment of shell closures at N=32,34. I. b decay of neutron-rich Sc isotopes
S. N. Liddick, P. F. Mantica, R. Broda, B. A. Brown, M. P. Carpenter, A. D. Davies, B. Fornal, T. Glasmacher, D. E. Groh, M. Honma, M. Horoi, R. V. F. Janssens, T. Mizusaki, D. J. Morrissey, A. C. Morton, W. F. Mueller, T. Otsuka, J. Pavan, H. Schatz, A. Stolz, S. L. Tabor, B. E. Tomlin, and M. Wiedeking National Superconducting Cyclotron Laboratory, Michigan State University, East Lansing, Mic...
متن کامل