Towards Multi-Document Summarization of Scientific Articles:Making Interesting Comparisons with SciSumm

Proceedings of the Workshop on Automatic Summarization for Different Genres, Media, and Languages, pages 8–15, Portland, Oregon, June 23, 2011. c⃝2011 Association for Computational Linguistics Towards Multi-Document Summarization of Scientific Articles:Making Interesting Comparisons with SciSumm Nitin Agarwal Language Technologies Institute Carnegie Mellon University nitina@cs.cmu.edu Ravi Shankar Reddy Language Technologies Resource Center IIIT-Hyderabad, India krs reddy@students.iiit.ac.in Kiran Gvr Language Technologies Resource Center IIIT-Hyderabad, India kiran gvr@students.iiit.ac.in Carolyn Penstein Ros´e Language Technologies Institute Carnegie Mellon University cprose@cs.cmu.edu Abstract We present a novel unsupervised approach to the problem of multi-document summariza- tion of scientific articles, in which the doc- ument collection is a list of papers cited to- gether within the same source article, other- wise known as a co-citation. At the heart of the approach is a topic based clustering of fragments extracted from each co-cited arti- cle and relevance ranking using a query gen- erated from the context surrounding the co- cited list of papers. This analysis enables the generation of an overview of common themes from the co-cited papers that relate to the con- text in which the co-citation was found. We present a system called SciSumm that em- bodies this approach and apply it to the 2008 ACL Anthology. We evaluate this summa- rization system for relevant content selection using gold standard summaries prepared on principle based guidelines. Evaluation with gold standard summaries demonstrates that our system performs better in content selec- tion than an existing summarization system (MEAD). We present a detailed summary of our findings and discuss possible directions for future research. 1 Introduction In this paper we present a novel, unsupervised ap- proach to multi-document summarization of scien- tific articles. While the field of multi-document sum- marization has achieved impressive results with col- lections of news articles, summarization of collec- tions of scientific articles is a strikingly different problem. Multi-document summarization of news articles amounts to synthesizing details about the same story as it has unfolded over a variety of re- ports, some of which contain redundant information. In contrast, each scientific article tells its own re- search story. Even with papers that address similar research questions, the argument being made is dif- ferent. Instead of collecting and arranging details into a single, synthesized story, the task is to abstract away from the specific details of individual papers and to find the common threads that unite them and make sense of the document collection as a whole. Another challenge with summarization of scien- tific literature becomes clear as one compares alter- native reviews of the same literature. Each review author brings their own unique perspective and ques- tions to bear in their reading and presentation of that literature. While this is true of other genres of doc- uments that have been the target of multi-document summarization work in the past, we don’t find query oriented approaches to multi-document summariza- tion of scientific articles. One contribution of this work is a technical approach to query oriented multi- document summarization of scientific articles that has been evaluated in …

Content truncated to 500 words