SciCorp: A Corpus of English Scientific Articles Annotated for Information Status Analysis

SciCorp: A Corpus of English Scientific Articles Annotated for Information Status Analysis Ina R¨osiger Institute for Natural Language Processing University of Stuttgart Germany ina.roesiger@ims.uni-stuttgart.de Abstract This paper presents SciCorp, a corpus of full-text English scientific papers of two disciplines, genetics and computational linguistics. The corpus comprises co-reference and bridging information as well as information status labels. Since SciCorp is annotated with both labels and the respective co-referent and bridging links, we believe it is a valuable resource for NLP researchers working on scientific articles or on applications such as co-reference resolution, bridging resolution or information status classification. The corpus has been reliably annotated by independent human coders with moderate inter-annotator agreement (average κ = 0.71). In total, we have annotated 14 full papers containing 61,045 tokens and marked 8,708 definite noun phrases. The paper describes in detail the annotation scheme as well as the resulting corpus. The corpus is available for download in two different formats: in an offset-based format and for the co-reference annotations in the widely-used, tabular CoNLL-2012 format. Keywords: Information Status, Co-reference, Bridging, Corpus Resource, Scientific Domain, English 1. Introduction Due to the rapid growth in publications, researchers are faced with the problem of having to extract relevant infor- mation out of an enormous number of journals, conference papers and workshop proceedings. Thus, the demand for a deeper, machine-driven understanding has gained interest in the research community, with applications such as sum- marisation, information extraction or citation analysis. As co-reference resolution and information status classifica- tion are important steps in the automatic processing of sci- entific papers, work on anaphora in scientific text is there- fore beneficial for these applications as well as for a better understanding of referring phenomena in this domain. Categorising the information status of noun phrases (NPs) or resolving anaphoric relations automatically re- quires annotated data for training and testing. However, for scientific text, annotated data are scarce: to the best of our knowledge, there is no full-text scientific corpus annotated with both information status and anaphoric relations. We thus present SciCorp, a scientific corpus of two dif- ferent disciplines, namely computational linguistics and ge- netics, for the purpose of information-structural analysis. Apart from standard predicted annotation layers the cor- pus features three types of manual annotation: co-reference clusters, bridging entities and their antecedents, and infor- mation status labels. We believe this corpus is useful for three tasks that are of interest to the NLP community: co- reference resolution, bridging resolution and information status classification. These tasks are briefly introduced in Section 2. We addressed the research question of resolving co- referent and associated references in scientific literature in R¨osiger and Teufel (2014). In this paper, an earlier version of the corpus was used as a basis for the experiments, but the corpus was not made publicly available as it was only annotated by one person. This paper presents details of the annotation process and describes the new corpus that was annotated by three independent annotators and that can be downloaded from our …

Content truncated to 500 words