Aspect based sentiment analysis for online reviews

Abstract

Learning good semantic vector representations for sentiment analysis in phrases, sentences and paragraphs is a challenging and ongoing area of natural language processing. In this paper, we propose a Convolution Neural Network for aspect level sentiment classification. Our model first builds a convolution neural network model to aspect extraction. Afterwards, we used a sequence labeling approach with Conditional Random Fields for the opinion target detection. Finally, we concatenate an aspect vector with every word embedding and apply a convolution neural network over it to determine the sentiment towards an aspect. Results of an experiment show that our method performs comparably well on Yelp reviews.