Discourse-Oriented Anaphora Resolution in Natural Language Understanding: A Review

To resolve various forms of definite reference u anaphora in particular early natural language understanding systems (reviewed in Hirst 1981) typically used a simple kind of history list of concepts previously mentioned in the input, with heuristics for selecting from this list. The history list was usually just a shift register containing the noun phrases from the last sentence or two, and the heuristics would take into account (among other things) selectional restrictions and syntactic constraints on pronominalization. SHRDLU (Winograd 1972) exemplifies this approach. Although able to resolve some types of reference, these systems were not able to handle reference in general, primarily because they did not take into account the effects of discourse structure on reference and pronominalization. This failure motivated work in computat ional discourse understanding that at tempted to exploit discourse structure, especially the relationship between reference and discourse theme, to resolve definite reference.