Extracting Specialized Translation Knowledge for Example-Based Machine Translation This talk provides an overview of example-based machine translation (EBMT) and contrast this approach to other machine translation paradigms. I present a number of EBMT systems and discuss my own work in the field. In particular, I present a program which extracts translation grammars from bilingual, parsed alignments and discuss a number of desirable properties for these grammars. Based on a large scale experiment, I show that these properties enhance translation quality when translating sublanguages.