Learning Structural Descriptions of Grammar Rules from Examples

Abstract

This . paper descnbes a LISP program that can learn English syntactic rules. The key idea is that the learn mg. cau be nude easy, g1vrn the right initial computational structure: syntactic knowledge is separated mto a fon·d mtrrprctcr and a variable set of h1ghly constramed pattern-action grammar rules. Only the gramma1 rulf's arc learned. via mduct1on from example sentences presented to . the program. The Interpreter is a s1mpllf1ed vers1on of Marcus's parser for English [1). which parses sentences without backup. The currently Implemented program acquires about 701. of a s1mplihed core grammar of English. What seems to makr the mductJon easy 1s that the rule structures and their actions are highly constrained: there are only four actions, ;~nd they manipulate only very local parts of the parse tree.