Resumen de Episteme

José Gabriel Amores Carredano , José Francisco Quesada Moreno

• This paper describes the overall architecture of Episteme, a tool for the development of efficient LFG-based MT systems following the classical transfer approach. The system incorporates a series of novel computational techniques which enhance the overall performance significantly: representation, storage and retrieval of very large feature structure-based knowledge bases, bidirectional event-driven bottom up parsing with top-down predictions and constructive unication with post-copy.