State-of-the-art Machine Translation (MT) does not perform well while translating sentiment components from source to target language. The components such as the sentiment holders, sentiment expressions and their corresponding objects and relations are not maintained during translation. In this paper, we described, how sentiment analysis can improve the translation quality by incorporating the roles of such components. We also demonstrated how a simple baseline phrase-based statistical MT (PB-SMT) system based on the sentiment components can achieve 33.88% relative improvement in BLEU for the under-resourced language pair English-Bengali.
展开▼
