Rapid Adaptive Development of Semantic Analysis Grammars

In this paper we describe a process for rapid development of semantic analysis grammars for Interlingual Machine Translation. The technique applies to existing systems and can be used to extend coverage into new languages quickly by separating the informant tasks performed by native speakers from the grammar writing tasks performed by engineers familiar with the system. A tool for automatic manipulation of parse trees uses information provided by both informant and engineer to create an example base of parse trees in the new language from which a grammar for that language can be read. Experimental results from a small-scale application of these tools are given to assess feasibility of the technique.