Position-aware Hybrid Attention Network for Aspect-Level Sentiment Analysis

Aspect-level sentiment analysis aims to predict the sentiment polarity of a given target in a review sentence. Most of the previous methods focus on capturing the context information of words across the sentence related to the target, ignoring the importance of the independent relationship between the opinion words and the target. To address this limitation, we propose a position-aware hybrid attention network model for aspect-level sentiment analysis, which incorporates not only the context information of words related to the target, but also the independent relationship between the opinion words related to the target. We conduct several comparable experiments on public laptop and restaurant datasets. The experimental results show that our proposed model achieves a more effective performance than the baseline models.