Generating Textual Summaries from the Bibliographies Contained in Scientific Literature

Abstract

This thesis explores the development and evaluation of a method for generating computer-generated textual summaries of bibliographies in scientific papers and conference proceedings. This method aims to enhance the utility of bibliographies by providing informative overviews that encapsulate the essence of cited sources. By integrating content analysis with metadata considerations, the method distills key information such as temporal trends, authorship patterns, research themes, and source characteristics into coherent summaries. This approach aids researchers in comprehending the contextual and thematic relevance of bibliographies and streamlines the research process by mitigating the need for exhaustive perusal of sources. The research investigates features and elements that bolster the efficacy of bibliography summaries, guided by the hypothesis that such summaries can aid scientists in their daily work. Through a mixed-methods evaluation involving qualitative interviews and a quantitative questionnaire, the study elucidates the preferences and requirements of academic professionals, revealing an inclination towards summaries that balance conciseness with comprehensive insight. Future avenues for this research encompass advanced topic analysis to unearth deeper thematic connections, incorporation of citation networks to elucidate scholarly discourse dynamics, and the development of customisable, interactive summary tools to cater to diverse user needs. Additionally, broadening user evaluations and integrating the summarisation method into existing research platforms are identified as critical steps towards enhancing the accessibility and impact of this method. Ultimately, this research posits that computer-generated bibliographic summaries hold the potential to transform literature review practices and foster more efficient scientific inquiry.