Knowledge-Enhanced Dual-Channel GCN for Aspect-Based Sentiment Analysis

As a subtask of sentiment analysis, aspect-based analysis (ABSA) refers to identifying the polarity given aspect. The state-of-the-art ABSA models are developed by using graph neural networks deal with semantics and syntax sentence. These methods challenged two issues. For one thing, semantic-based convolution fail capture relation between aspect its opinion word. another, minor attention is assigned word within convolution, resulting in introduction contextual noise. In this work, we propose knowledge-enhanced dual-channel convolutional network. On task ABSA, netwok (GCN) syntactic-based GCN established. With respect semantic learning, sentence enhanced commonsense knowledge. multi-head mechanism taken construct filter noise, which facilitates information aggregation words. syntactic processing, dependency tree pruned remove irrelevant words, based on more weights Experiments carried out four benchmark datasets evaluate working performance proposed model. Our model significantly outperforms baseline verifies effectiveness tasks.