| Summary: | This paper describes in detail the computational process of finding the non-indexical linguistic objects to which pronoun anaphoras resolve. The successful development of discrete steps for anaphora resolution is critical for any natural language interface to a variety of systems. The types of information available for resolution are first identified and the past approaches to resolution for each type are critically reviewed. This paper then describes how anaphora resolution is implemented in the present system using all types of information, including the discourse information of focus, background, and focus shift.
|
|---|