User Bias Removal in Fine Grained Sentiment Analysis.
Rahul Wadbude, Vivek Gupta, Dheeraj Mekala, Janish Jindal, Harish Karnick · arXiv (Cornell University) · 2016
Fine-grained sentiment analysis of text reviews has recently gained a lot of attention in the natural language processing community. Most work done earlier focuses on creating efficient feature representations of text reviews for classification. The methods generally ignore other common attributes like user identity, product identity, and helpfulness rating during fine-grained sentiment classification. A major problem in current classification models is noise due to the presence of user bias in review ratings. We propose two simple statistical methods to remove such noise and improve fine-grained sentiment classification. We apply our methods on the SNAP published Amazon Fine Food Reviews dataset and on two major categories Electronics and Movies and TV of the e-Commerce Reviews data-set. After removing user bias, we get improved fine-grained sentiment classification with three commonly used feature representations.