Paper: A Tree-to-String Phrase-based Model for Statistical Machine Translation

ACL ID W08-2119
Title A Tree-to-String Phrase-based Model for Statistical Machine Translation
Venue International Conference on Computational Natural Language Learning
Session Main Conference
Year 2008
Authors

Thoughphrase-basedSMThas achieved high translationquality, it still lacks of generaliza- tion abilityto captureword order differences betweenlanguages.In this paperwe describe a general method for tree-to-stringphrase- based SMT. We study how syntactic trans- formation is incorporatedinto phrase-based SMTand its effectiveness. We designsyntac- tic transformationmodelsusingunlexicalized form of synchronouscontext-free grammars. These models can be learned from source- parsedbitext. Our systemcan naturallymake use of both constituentand non-constituent phrasaltranslationsin thedecodingphase.We consideredvarious levels of syntacticanaly- sis ranging from chunking to full parsing. Our experimentalresultsof English-Japanese and English-Vietnamese translation showed a significantimprovement over...