Aspect-Based Sentiment Analysis of Nuclear Energy Tweets with Attentive Deep Neural Network

Abstract

Opinion mining of social networking sites like Facebook and Twitter plays an important role in exploring valuable online user-generated contents. In contrast to sentence-level sentiment classification, the aspect-based analysis which can infer polarities towards various aspects in one sentence could obtain more in-depth insight. However, in traditional machine learning approaches, training such a fine-grained model often needs certain manual feature engineering. In this article, we proposed a deep learning model for aspect-level sentiment analysis and applied it to nuclear energy related tweets for understanding public opinions towards nuclear energy. We also built a new dataset for this task and the evaluation results showed that our attentive neural network could obtain insightful inference in rather complex expression forms and achieve state-of-the-art performance.