Learning what the crowd can do: A case study on focus annotation

Abstract

This paper addresses the question of how to explore and advance the conceptualization and applicability of information structural notions to support the analysis of authentic data. With this we aim at further establishing where advances in linguistic modeling also result in quantifiable gains in real-life tasks. Can, for example, computational linguistic applications be improved by integrating information structural notions? One of the necessary prerequisites for answering this question are large enough sets of data which are annotated with the relevant information structural concepts. The main problem here is that notions like focus are often discussed in theoretic literature by means of example sentences but rarely analyzed in substantial amounts of authentic data.