Text this: Corpus-based and computational approaches to discourse anaphora