Abstract
A major problem in the development of models of language understanding is the manner in which a reader extracts more from a text than the meaning which its individual sentences convey when interpreted in isolation. This paper addresses a major aspect of this problem: the creation of a model for the proper resolution of “contextually-induced” references (with “reference” defined much more broadly than is usually the case). A unified view of contextual effects on reference is presented, and the procedures (algorithms) required for reference resolution are analyzed. A framework for the representation of context is proposed, based on a small set of intersentential relations that represent the manner in which a text structures the concepts it presents, rather than the a priori semantic and pragmatic relations between these concepts. A top-down procedural model is proposed for the dynamic extraction and use of context structure to resolve references.
| Original language | English |
|---|---|
| Pages (from-to) | 25-71 |
| Number of pages | 47 |
| Journal | Discourse Processes |
| Volume | 3 |
| Issue number | 1 |
| DOIs | |
| State | Published - 1 Jan 1980 |