Skip to Main Content
Current statistical speech translation approaches predominantly rely on just text transcripts and are limited in their use of rich contextual information such as prosody and discourse function. In this paper, we explore the role of discourse context characterized through dialog acts (DAs) in statistical translation. We present a bag-of-words (BOW) model that exploits DA tags in translation and contrast it with a phrase table interpolation approach presented in previous work. In addition to producing interpretable DA-annotated target language translations through our framework, we also obtain consistent improvements in terms of automatic evaluation metrics such as lexical selection accuracy and BLEU score using both the models. We also analyze the performance improvements per DA tag. Our experiments indicate that questions, acknowledgments, agreements and appreciations contribute to more improvement in comparison to statements.