Automatically Identifying Implicit Arguments to Improve Argument Linking and Coherence Modeling
Michael Roth, Anette Frank · 2013
Implicit arguments are a discourse-level phenomenon that has not been extensively studied in semantic processing. One reason for this lies in the scarce amount of annotated data sets available. We argue that more data of this kind would be helpful to improve existing approaches to linking implicit arguments in discourse and to enable more in-depth studies of the phenomenon itself. In this paper, we present a range of studies that empirically validate this claim. Our contributions are threefold: we present a heuristic approach to automatically identify implicit arguments and their antecedents by exploiting comparable texts; we show how the induced data can be used as training data for improving existing argument linking models; finally, we present a novel approach to modeling local coherence that extends previous approaches by taking into account non-explicit entity references. 1