Generating Impact-Based Summaries for Scientific Literature

Proceedings of ACL-08: HLT, pages 816–824, Columbus, Ohio, USA, June 2008. c⃝2008 Association for Computational Linguistics Generating Impact-Based Summaries for Scientific Literature Qiaozhu Mei University of Illinois at Urbana- Champaign qmei2@uiuc.edu ChengXiang Zhai University of Illinois at Urbana- Champaign czhai@cs.uiuc.edu Abstract In this paper, we present a study of a novel summarization problem, i.e., summarizing the impact of a scientific publication. Given a pa- per and its citation context, we study how to extract sentences that can represent the most influential content of the paper. We propose language modeling methods for solving this problem, and study how to incorporate fea- tures such as authority and proximity to ac- curately estimate the impact language model. Experiment results on a SIGIR publication collection show that the proposed methods are effective for generating impact-based sum- maries. 1 Introduction The volume of scientific literature has been growing rapidly. From recent statistics, each year 400,000 new citations are added to MEDLINE, the major biomedical literature database 1. This fast growth of literature makes it difficult for researchers, espe- cially beginning researchers, to keep track of the re- search trends and find high impact papers on unfa- miliar topics. Impact factors (Kaplan and Nelson, 2000) are useful, but they are just numerical values, so they cannot tell researchers which aspects of a paper are influential. On the other hand, a regular content- based summary (e.g., the abstract or conclusion sec- tion of a paper or an automatically generated topical summary (Giles et al., 1998)) can help a user know 1http://www.nlm.nih.gov/bsd/history/tsld024.htm about the main content of a paper, but not necessar- ily the most influential content of the paper. Indeed, the abstract of a paper mostly reflects the expected impact of the paper as perceived by the author(s), which could significantly deviate from the actual im- pact of the paper in the research community. More- over, the impact of a paper changes over time due to the evolution and progress of research in a field. For example, an algorithm published a decade ago may be no longer the state of the art, but the problem def- inition in the same paper can be still well accepted. Although much work has been done on text sum- marization (See Section 6 for a detailed survey), to the best of our knowledge, the problem of impact summarization has not been studied before. In this paper, we study this novel summarization problem and propose language modeling-based approaches to solving the problem. By definition, the impact of a paper has to be judged based on the consent of research community, especially by people who cited it. Thus in order to generate an impact-based sum- mary, we must use not only the original content, but also the descriptions of that paper provided in papers which cited it, making it a challenging task and dif- ferent from a regular summarization setup such as news summarization. Indeed, unlike a regular sum- marization system which identifies and interprets the topic of a document, an impact summarization …

Content truncated to 500 words