Paper: Joint Morphological-Lexical Language Modeling for Machine Translation

ACL ID N07-2037
Title Joint Morphological-Lexical Language Modeling for Machine Translation
Venue Human Language Technologies
Session Short Paper
Year 2007
Authors

We present a joint morphological-lexical language model (JMLLM) for use in statistical machine trans- lation (SMT) of language pairs where one or both of the languages are morphologically rich. The pro- posed JMLLM takes advantage of the rich morphol- ogy to reduce the Out-Of-Vocabulary (OOV) rate, while keeping the predictive power of the whole words. It also allows incorporation of additional available semantic, syntactic and linguistic informa- tion about the morphemes and words into the lan- guage model. Preliminary experiments with an English to Dialectal-Arabic SMT system demon- strate improved translation performance over trigram based baseline language model.