Pattern‐matching and inference in story understanding∗
Richard E. Cullingford · Discourse Processes · 1979
A crucial issue in story understanding is recognizing and using the context in which the problems which make a story “interesting” are worked out. This paper discusses a computer text‐processing system, SAM, which uses Scripts to provide such context when reading newspaper articles referring to relatively stereotyped situations such as car accidents, state visits, train wrecks, etc. In SAM, a method of pattern‐matching is used, first to decide to which situation a story refers, then to follow the story through the context. The patterns which make up a Script are augmented by a set of inference processes which help establish the low‐level connections which tie the sentences of a text together. Since such inferences are needed in all types of story understanding, SAM's pattern‐match‐and‐inference cycle provides a model for one type of processing any automatic understander will have to do.