Aspect-Oriented Sentiment Classification using BiGRU-CNN model

Abstract

People on the Internet have generated a large amount of commentary data to share their opinions about products and services in their daily lives which include large commercial value. For these comment sentences, they often include several comment aspects, and the sentiment varies on these aspects, making the overall meaning of the sentence meaningless for polarization. The purpose of the aspect-level sentiment classification is to recognize target’s sense extremity in context. Deep Learning is evolving in an increasingly mature direction, and the utilization of deep learning methods to detect emotion has become increasingly popular. A sentiment classification model is propsoed by combining a convolutional neural network and a bidirectional gated recurrent unit.Bidirectional gated recurrent unit is similar to Long short-term memory, a time cyclic neural network with a lesser processing complexity. The model first extracts sequence features of the text through the bidirectional gated recurrent unit and then extracts the local static features of the text through the convolutional neural network. Finally, the Sigmoid classifier is used for the final sentiment classification.