ABSTRACT
Sentiment analysis is extremely useful in social media monitoring as it allows us to gain an overview of the wider public opinion behind certain topics. While sentiment analysis of European languages has been widely studied and experimented on, such tools for Indian languages, are scarce and don’t provide much value as accuracy scores are not acceptable. In this paper, we tried to mitigate this problem by using Machine Translation. The main idea was to train a sentiment analysis system for English that provides high accuracy scores. This system will then be used to annotate the test English sentences with thei respective sentiment labels. Also, a machine translation system, trained to translate the English language to Bengali, was used to translate the same test sentences into Bengali. As a result, we were able to develop a labeled Bengali sentiment corpus. Thereafter, we were able to train a Bengali sentiment analysis system, using this corpus. When tested using automated and manual metrics, the developed model garnered very good qualitative scores.
