A Survey Paper on Cross-Domain Sentiment Analysis

Abstract: The Internet contains important information on its user’s opinions and sentiments. The extraction of those unstructured data is known as opinion mining and sentiment analysis. Basically Sentiment classification aims to automatically predict sentiment polarity (e.g., positive or negative) of users publishing sentiment data (e.g., reviews, blogs). This paper presents a short survey on cross-domain sentiment analysis techniques suggested and implemented recently. These techniques are then compared on the basis of feature expansion, number of source domain used, labeled and unlabeled data etc. These are then summarised and analysed. General challenges for performing cross-domain sentiment analysis are also discussed.

Keywords: cross-domain sentiment classification, support vector machine, domain adaptation, SentiWordNet.