Sentiment Polarity in Translation

Abstract

Previous year, many researchers have been sentiment analysis on many focus languages. They analyzed and categorizing opinions expressed in a text. People express their opinions and feeling on social media as a daily routine. For sentiment analysis work, data plays an important role. Thus, social media become interested platform for opinion mining. On the other hand, low resource languages face less of sentiment resources (such as sentiment lexicon, corpus) than English language. It is needed to overcome language barriers and realize a sentiment platform capable of scoring in different languages when global opinion is need to decide something. In this paper, the expectations and limitations of machine translation in sentiment polarity task for Myanmar language is presented. We experiment with comments of particular news and general news that are expressed in social media news pages. Results show that sentiment transfer can be successful through human translation. This also demonstrates that translation from Myanmar to English has a significant eff ect on the preservation of sentiment by using translation engine. This happens primarily due to nature of Language but the results show that machine translation quality plays the important role in this work.