Electronic Thesis and Dissertation Repository


Doctor of Philosophy


Computer Science


Robert E. Mercer


In recent years, there has been a dramatic increase in scientific publications across the globe. To help navigate this overabundance of information, methods have been devised to find papers with related content, but they are lacking in the ability to provide specific information that a researcher may need without having to read hundreds of linked papers. The search and browsing capabilities of online domain specific scientific repositories are limited to finding a paper citing other papers, but do not point to the specific text that is being cited. Providing this capability to the research community will be beneficial in terms of the time required to acquire the amount of background information they need to undertake their research. In this thesis, we present our effort to develop a citation linkage framework for finding those sentences in a cited article that are the focus of a citation in a citing paper. This undertaking has involved the construction of datasets and corpora that are required to build models for focused information extraction, text classification and information retrieval. As the first part of this thesis, two preprocessing steps that are deemed to assist with the citation linkage task are explored: method mention extraction and rhetorical categorization of scientific discourse. In the second part of this thesis, two methodologies for achieving the citation linkage goal are investigated. Firstly, regression techniques have been used to predict the degree of similarity between citation sentences and their equivalent target sentences with medium Pearson correlation score between predicted and expected values. The resulting learning models are then used to rank sentences in the cited paper based on their predicted scores. Secondly, search engine-like retrieval techniques have been used to rank sentences in the cited paper based on the words contained in the citation sentence. Our experiments show that it is possible to find the set of sentences that a citation refers to in a cited paper with reasonable performance. Possible applications of this work include: creation of better science paper repository navigation tools, development of scientific argumentation across research articles, and multi-document summarization of science articles.