Argumentation in scholarly biomedical literature: A computational theory and its software implementation

点击翻译按钮获取中文摘要

Scientific research articles are written by scientists to argue for a new scientific claim and to persuade other scientists to accept this claim. The philosopher Toulmin has proposed a framework that captures how people use language to convince others. This framework, drawn from observation and analysis of legal arguments and transformed into the experimental science arena, is composed of a scientific claim backed up with evidence drawn from the authors' scientific experiments. Far too much new scientific knowledge is being accumulated in the biomedical sciences for any person to keep abreast of. New computer technologies can help experimental scientists and clinicians cope with this deluge of information by aggregating all of this new knowledge into a useable form.  My research program will provide some of these new computational technologies. A key to these new technologies will be methods to capture these persuasive arguments in a format that is manipulable by computers. Argumentation is accomplished through the use of a variety of rhetorical devices, all having been well-studied non-computationally by applied linguists and rhetoricians of science: rhetorical methods in (oral and written) language, and the use of citations, figures, tables, etc. in scientific writing. My research program will capture these rhetorical devices that are found specifically in published biomedical research papers in a computational theory of argumentation structures and translate these text-based argumentation structures into Abstract Dialectical Frameworks, a state-of-the-art Artificial Intelligence model of argumentation. To construct the argumentation structure, these components must be identified. This is accomplished using standard supervised Machine Learning methodology which requires analysis of the text using Computational Linguistics techniques to extract text-based features to generate the machine learned model. The text analysis will use the many widely used information resources (ontologies, databases, software tools) that have been developed by Biomedical and Computational Linguistics researchers for this purpose. In addition we will also develop novel methods to extract information from tables. Outcomes of this research program will have profound societal benefits in the fields of experimental and clinical medicine. Having computer-based representations of scientific arguments can lead to significant changes in how biomedical research is performed. Inconsistent experimental results can be detected, unrecognized biomedical relationships can be discovered, and the vast biomedical literature will be synthesized into a more accessible form for clinicians' use. Clinicians, with this information, will be able to diagnose rare cases more confidently and as part of their clinical practice will be able to choose more appropriate genetic tests to confirm their diagnoses.**