An Efficient Insertion-Only Error-Corrector for LL(1) Parsers

Abstract

An LL(1)-based error-corrector which operates by insertion-only is studied. The corrector is able to correct and parse any input string. It is efficient (linear in space and time requirements) and chooses least-cost insertions (as defined by the user) in correcting syntax errors. Moreover, the error-corrector can be generated automatically from the grammar and a table of terminal symbol insertion costs. The class of LL(1) grammars correctable by this method contains (with minor modifications) grammars used to specify most common programming languages. Preliminary results suggest that this method can be used to advantage in LL(1)-driven compilers.