Resolving connective ambiguity
Automatic discourse parsing refers to the identification of coherence relationsand deriving a structural description for a text. Such parsers can derive muchinformation from the presence of surface cues, especially connectives. Theselexical signals, however, are ambiguous: Many have additional, non-connectivereadings; also, many connectives can signal more than one coherence relation.In this paper, we discuss the first problem, focusing on English and German:How many connectives are ambiguous, and how frequent are these in the twolanguages? Then we examine computational approaches for resolving suchambiguities. For English, we provide an overview of relevant work by otherresearchers, while for German we largely present our own studies on the utilityof part-of-speech tagging for connective disambiguation.