Combining dynamic local context focus and dependency cluster attention for aspect-level sentiment classification.

Aspect-level sentiment classification (ASC), as a subtask of Aspect-based sentiment analysis (ABSA), aims to analyze the sentiment polarity of different aspect terms in a sentence. Although the existing methods have been proven to be effective, they fail to effectively identify the range of local context and fully leverage the nonequivalent property of dependency relation. Hence, we propose a concept of dependency cluster and design two modules named Dynamic Local Context Focus (DLCF) and Dependency Cluster Attention (DCA) respectively. The DLCF can dynamically capture the range of local context based on the different max distance from the target aspect term to its context words and the DCA allows the model to pay more attention to the cluster which is more critical for sentiment classification. Together with the two modules, we then propose the DLCF-DCA model, in which the DLCF is equipped before DCA. Considering the DLCF has masked or weighted down the less-semantic-relative words, the semantic information can therefore be better extracted in DCA. Experiments conducted on six benchmark datasets demonstrate that DLCF-DCA achieves the state-of-the-art results. Moreover, the ablation experiment results also verify the effectiveness of each part in DLCF-DCA. [ABSTRACT FROM AUTHOR]