A Three-Way Perspective on Scientific Discourse Annotation for Knowledge Extraction

摘要

This paper presents a three-way perspective on the annotation of discourse in scientific literature. We use three different schemes, each of which focusses on different aspects of discourse in scientific articles, to annotate a corpus of three full-text papers, and compare the results. One scheme seeks to identify the core components of scientific investigations at the sentence level, a second annotates metaknowledge pertaining to bio-events and a third considers how epistemic knowledge is conveyed at the clause level. We present our analysis of the comparison, and a discussion of the contributions of each scheme.