Construction of a hierarchical translation memory

Translation memories are promising devices for automatic translation. Their main weakness, however, is poor coverage on unseen text. In this paper, the use of a hierarchical translation memory, consisting of a cascade of finite state transducers, is proposed. A number of transducers is applied to convert sentence pairs from a bilingual corpus into translation patterns, which are then used as a translation memory. Preliminary results on the German English VERBMOBIL corpus are given.