Syntactically Enhanced Dependency-POS Weighted Graph Convolutional Network for Aspect-Based Sentiment Analysis

Aspect-based sentiment analysis (ABSA) is a fine-grained task of that presents great benefits to real-word applications. Recently, the methods utilizing graph neural networks over dependency trees are popular, but most them merely considered if there exist dependencies between words, ignoring types these dependencies, which carry important information, as with different have effects. In addition, they neglected correlations and part-of-speech (POS) labels, helpful for imformation. To address such limitations deficiency insufficient syntactic semantic feature mining, we propose novel model containing three modules, aims leverage more reasonably by distinguishing extracting beneficial features further enhance performance. enrich word embeddings, design encoder (SynFE). particular, Dependency-POS Weighted Graph Convolutional Network (DPGCN) weight attention mechanism proposed. Additionally, capture aspect-oriented extractor (SemFE). Extensive experiments on five popular benchmark databases validate our can better employ information effectively extract favorable achieve new state-of-the-art