We compare the potential of two classes of linear and hierarchical models of discourse to determine co-reference links and resolve anaphors. The comparison uses a corpus of thirty texts, which were manually annotated for co-reference and discourse structure.
展开▼