Augmenting Manual Dictionaries for Statistical Machine Translation Systems
Stephan Vogel, Christian Monson
Language Technologies Institute, Carnegie Mellon University
We show that the usefulness of manually created dictionaries can be enhanced for a statistical machine translation system when new translations are automatically added which are simple morphological transformations (plural forms, different verb inflections) of the original. Further improvement is possible when assigning probabilities to the lexicon entries. We describe a method to do this on the basis of an automatically trained statistical lexicon. Experimental results are given for Chinese to English translation tasks and show a significant improvement in translation quality.
Dictionary, statistical machine translation