Detecting Grammar Errors in Children's Writing: A Finite State Approach

This paper reports on the development of a finite state system for finding grammar errors in Swedish text written by children. The writing problems are more frequent for this group and the distribution of the error types is different from texts written by adults. The detection approach involves subtraction of automata that represent two “positive” grammars with varying degree of detail. The difference between the automata corresponds to the search for writing problems that violate the grammars. The constituents of various fragments are identified by the use of lexical-prefix-first strategy and application of incremental parsing. While the grammatical coverage is still rather small, yielding rather large numbers of false alarms, the technique can be applied to agreement phenomena, verb selection phenomena and some word order phenomena. The aim is to include also detection of missing sentence boundaries.