Emotion Detection from Text Based Document and classification of Cross-Domain Sentiment

Sentiment analysis focuses on analyzing web documents, especially user-generated content such as product reviews, to identify opinionated documents, sentences and opinion holders. Most of the time classifiers trained in one domain do not perform well in another domain. The existing approaches do not detect sentiment and topics simultaneously. Sentiments may differ with topics. Our proposed model called Joint Sentiment Topic (JST) model to detect sentiments and topics simultaneously from text. This model is based on Gibbs sampling algorithm. Besides, unlike supervised approaches to opinion mining which often fail to produce good performance when shifting to other domains, the semi-supervised nature of JST makes it highly portable to other domains. JST model performs better when compared to existing supervised approaches.