Annotating evidence-based argumentation in biomedical text

A new challenge in natural language processing, argumentation mining is the automatic identification of an argument's premises, conclusion, and argumentation scheme, and relationships between arguments. Argumentation mining could provide critical context for information extraction and question answering and support new forms of summarization and citation indexing. However, argumentation mining, in this sense, has not yet been addressed in BioNLP. This theoretical paper contributes towards annotation of argumentation in biomedical/biological corpora. We show that argumentative zone models and models of discourse coherence do not represent the same aspects of discourse as a model of evidence-based argumentation. We explore the challenges of annotation of argumentation in full-text biomedical articles and describe the steps we have taken towards annotating a biomedical corpus for argumentation mining research.