Dual-Channel Edge-Featured Graph Attention Networks for Aspect-Based Sentiment Analysis

The goal of aspect-based sentiment analysis (ABSA) is to identify the polarity specific aspects in a context. Recently, graph neural networks have employed dependent tree syntactic information assess link between and contextual words; nevertheless, most this research has neglected phrases that are insensitive effect various sentence. In paper, we propose dual-channel edge-featured attention model (AS-EGAT), which builds an aspect by enhancing dependency representation key words mutual affective relationship context semantic through self-attention mechanism. We use edge features as significant factor determine weight coefficient mechanism efficiently mine (GAT). As result, can connect important related when dealing with lack obvious expressions, pay close word multiple-word aspects, extract from sentences not sensitive trees looking at features. Experimental results show our proposed AS-EGAT superior current state-of-the-art baselines. Compared baseline models LAP14, REST15, REST16, MAMS, T-shirt, Television datasets, accuracy increased 0.76%, 0.29%, 0.05%, 0.15%, 0.22%, 0.38%, respectively. macro-f1 score 1.16%, 1.23%, 0.37%, 0.53%, 1.93%