Argumentative structure of scientific abstracts
Description
In order to explore the possibility of leveraging discourse information for the identiffication of argumentative components and relations we add a new annotation layer to a subset of the Discourse Dependency TreeBank for Scientiffic Abstracts (SciDTB). [1] We introduce a ffine-grained annotation schema aimed at capturing information that accounts for the specifficities of the scientiffic discourse, including the type of evidence that is offered to support a statement (e.g., background information, experimental data or interpretation of results). [1] Yang, A., Li, S.: SciDTB: Discourse dependency TreeBank for scientiffc abstracts. In: Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (ACL 2018) (Volume 2: Short Papers). pp. 444{449. Association for Computational Linguistics, Melbourne, Australia (Jul 2018)