Knowledge Adaptation for Cross-Domain Opinion Mining

Automatic opinion mining of web 2.0 texts is critical for understanding people's viewpoints and assisting them in making informed decisions. Trained machines perform well in the same domain to predict the sentiment polarity but performance decreases drastically when the same machine is applied directly to other domains. Creating a labeled data for every field is an expensive and inefficient procedure. We introduce a framework to determine the domain-independent words in both domains by employing feature extraction techniques to bridge the gap across the domains. To train a classifier and analyze the sentiment polarity of the target domain, we employed these features. The experimental results are compared with different existing state-of-art approaches and evaluate the execution and effectiveness of the suggested framework.