Coherent Citation-Based Summarization of Scientific Papers

Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics, pages 500–509, Portland, Oregon, June 19-24, 2011. c⃝2011 Association for Computational Linguistics Coherent Citation-Based Summarization of Scientific Papers Amjad Abu-Jbara EECS Department University of Michigan Ann Arbor, MI, USA amjbara@umich.edu Dragomir Radev EECS Department and School of Information University of Michigan Ann Arbor, MI, USA radev@umich.edu Abstract In citation-based summarization, text written by several researchers is leveraged to identify the important aspects of a target paper. Previ- ous work on this problem focused almost ex- clusively on its extraction aspect (i.e. selecting a representative set of citation sentences that highlight the contribution of the target paper). Meanwhile, the fluency of the produced sum- maries has been mostly ignored. For exam- ple, diversity, readability, cohesion, and order- ing of the sentences included in the summary have not been thoroughly considered. This re- sulted in noisy and confusing summaries. In this work, we present an approach for produc- ing readable and cohesive citation-based sum- maries. Our experiments show that the pro- posed approach outperforms several baselines in terms of both extraction quality and fluency. 1 Introduction Scientific research is a cumulative activity. The work of downstream researchers depends on access to upstream discoveries. The footnotes, end notes, or reference lists within research articles make this accumulation possible. When a reference appears in a scientific paper, it is often accompanied by a span of text describing the work being cited. We name the sentence that contains an explicit reference to another paper citation sentence. Cita- tion sentences usually highlight the most important aspects of the cited paper such as the research prob- lem it addresses, the method it proposes, the good results it reports, and even its drawbacks and limita- tions. By aggregating all the citation sentences that cite a paper, we have a rich source of information about it. This information is valuable because human ex- perts have put their efforts to read the paper and sum- marize its important contributions. One way to make use of these sentences is creat- ing a summary of the target paper. This summary is different from the abstract or a summary gener- ated from the paper itself. While the abstract rep- resents the author’s point of view, the citation sum- mary is the summation of multiple scholars’ view- points. The task of summarizing a scientific paper using its set of citation sentences is called citation- based summarization. There has been previous work done on citation- based summarization (Nanba et al., 2000; Elkiss et al., 2008; Qazvinian and Radev, 2008; Mei and Zhai, 2008; Mohammad et al., 2009). Previous work fo- cused on the extraction aspect; i.e. analyzing the collection of citation sentences and selecting a rep- resentative subset that covers the main aspects of the paper. The cohesion and the readability of the pro- duced summaries have been mostly ignored. This resulted in noisy and confusing summaries. In this work, we focus on the coherence and read- ability aspects of the problem. Our …

Content truncated to 500 words