GRIND‐1: First report on the magic grinder story comprehension project∗
Chuck Rieger · Discourse Processes · 1978
GRIND‐1, a story comprehension model inspired by the Walt Disney children's story The Magic Grinder, makes its debut. The paper outlines some goals for any story comprehension model, then describes our specific goals in the end‐to‐end comprehension of this 40‐page children's story. This first GRIND‐1 report mentions the six major model aspects (language analysis and meaning representation, plot representation, the character personality trait constructer, the two‐character relationship modeler, the visual parser, and the prediction‐fulfillment mechanism), then focuses on GRIND‐1's theory of meaning based language analysis. This theory of language, the Sense Net Parser, rejects the usual approach in which a central system of rules prevails. Instead, GRIND‐1's knowledge about language is distributed throughout a system of word experts called “sense selection nets” that, when run in cooperative group environments, construct meaning representations for sentences in context. Annotated traces of the language‐related parts of the system for the first part of the story are included.