Purpose and Polarity of Citation: Towards NLP-based Bibliometrics

Proceedings of NAACL-HLT 2013, pages 596–606, Atlanta, Georgia, 9–14 June 2013. c⃝2013 Association for Computational Linguistics Purpose and Polarity of Citation: Towards NLP-based Bibliometrics Amjad Abu-Jbara Department of EECS University of Michigan Ann Arbor, MI, USA amjbara@umich.edu Jefferson Ezra Department of EECS University of Michigan Ann Arbor, MI, USA jezra@umich.edu Dragomir Radev Department of EECS and School of Information University of Michigan Ann Arbor, MI, USA radev@umich.edu Abstract Bibliometric measures are commonly used to estimate the popularity and the impact of pub- lished research. Existing bibliometric mea- sures provide “quantitative” indicators of how good a published paper is. This does not nec- essarily reflect the “quality” of the work pre- sented in the paper. For example, when h- index is computed for a researcher, all incom- ing citations are treated equally, ignoring the fact that some of these citations might be neg- ative. In this paper, we propose using NLP to add a “qualitative” aspect to biblometrics. We analyze the text that accompanies citations in scientific articles (which we term citation context). We propose supervised methods for identifying citation text and analyzing it to de- termine the purpose (i.e. author intention) and the polarity (i.e. author sentiment) of citation. 1 Introduction An objective and fair evaluation of the impact of published research requires both quantitative and qualitative assessment. Existing bibliometric measures such as H-Index (Hirsch, 2005; Hirsch, 2010), G-index (Egghe, 2006), and Impact Fac- tor (Garfield, 1994) focus on the quantitative aspect of this evaluation which dose not always correlate with the qualitative aspect. For example, the number of papers published by a researcher only tells how productive she or he is. It does not say anything about the quality or the im- pact of the work. Similarly, the number of citations that a paper receives should not be used to gauge the quality of the work as it really only measures the popularity of the work and the interest of other researchers in it (Garfield, 1979). Controversial pa- pers or those based on fabricated data or experiments may receive a large number of citations. A popular example of fraudulent research that deceived many researchers and caught media attention was the case of a South Korean research scientist, Hwang Woo- suk, who was found to have faked his research re- sults in the area of human stem cell cloning. His re- search was published in Science and received close to 200 citations after the fraud was discovered. The vast majority of those citations were negative. This suggests that the purpose of citation should be taken into consideration when biblometric mea- sures are computed. Negative citations should be weighted less than positive or neutral citations. This motivates the need to automatically distinguish be- tween positive, negative, and neutral citations and to identify the purpose of a citation; i.e. the author’s in- tention behind choosing a published article and cit- ing it. This analysis of citation purpose and polarity can be useful for many applications. For example, it can be used to …

Content truncated to 500 words