A Citation Centric Annotation Scheme for Scientific Articles

A Citation Centric Annotation Scheme for Scientific Articles Angrosh M.A. Stephen Cranefield Nigel Stanger Department of Information Science, University of Otago, Dunedin, New Zealand (angrosh, scranefield, nstanger}@infoscience.otago.ac.nz Abstract This paper presents an annotation scheme for modelling citation contexts in scientific articles. We present an argumentation framework based on the Toulmin model for scientific articles and develop an annotation scheme with different context types based on the argumentation model. We present the results of the inter-rater reliability study carried out for studying the reliability of our annotation scheme. 1 Introduction Citations play an important role in scientific writing. However, there are not many tools that provide citation context based information services. The citation services provided by academic search engines such as Google Scholar1 includes information about the number of citing documents and links to citing articles. Search can also be performed for keywords in citing articles. Citation focused tools such as CiteSeerX 2 and Microsoft Academic Search 3 engines go a little further in identifying the passage of citations in citing articles. The objective of such services is to facilitate quick access to citation content to aid the learning process. However, the high volume of research content renders it difficult to achieve optimum use of these services. Identifying this need, we proposed to develop tools for providing intelligent citation context based information services. However, an annotation scheme for citation contexts is one of the key requirements of citation context based information tools. An annotation scheme providing citation contexts can help classifying citation passages and provide betterin citation context based information services. Accordingly, we studied the existing annotation 1 http://scholar.google..com 2 http://citeseerx.ist.psu.edu 3 http://academic.research.microsoft.com/ schemes and noted that it was difficult to use these schemes for our application and proposed to develop a new annotation scheme. Angrosh, Cranefield and Stanger (2012a) have successfully used the annotation scheme with machine learning techniques for developing intelligent citation context based tools. These included a linked data application (Angrosh, Cranefield, and Stanger, 2012b) and a Web-based application4. We present in this paper our annotation scheme designed to represent citation contexts based on an argumentation model, which can be used to develop citation context based information tools. 2 Related Work Over the years, several proposed annotation schemesresearchers have for scientific articles. These schemes can be classified into two categories: (a) those that consider the full text of an article; and (b) those addressing citation sentences only. 2.1 Annotation Schemes for Full Text Conceptualizing the idea of ‘argumentative zoning’, Teufel (1999) proposed an annotation scheme of seven categories and called them argumentative zones for sentences. Mizuta and Collier (2004a) extended Teufel’s argumentation scheme (Teufel, 1999) for zone analysis in biology texts and provided a scheme of seven categories. Langer et al. (2004) noted that newer applications in areas such as the Semantic Web required richer and more fine-grained annotation of seven topic types for documents. Motivated by the need to identify passages of reliable scientific facts, Wilbur et al. (2006) devised an annotation scheme of five categories for …

Content truncated to 500 words