Abstract: Aspect-based sentiment analysis aims to infer the sentiment polarity of specific aspects. Existing methods often adopt an equal-weight strategy when aggregating neighbor node information, failing to distinguish the differences in node importance, which leads to insufficient semantic-syntactic coordination. Graph attention networks rely on indirect syntactic associations to capture the relationship between aspect words and sentiment words, resulting in a lack of global features and low distinguishability in the sentiment representation space, thereby causing ambiguous sentiment space. To address the above issues, this paper proposes a graph convolutional network model inte-grating mutual information alignment and supervised contrastive learning, denoted as MIC-GCN (Mutual Information and Con-trastive Learning-based Graph Convolutional Network Model).. The model first converts text words into vector representations through an embedding layer. Then, a dual-attention dynamic routing module is designed to combine the dynamic interaction of self-attention and aspect-aware attention to generate an attention score matrix, and a dynamic masking matrix is constructed based on syntactic distance to optimize the utilization of structural information. Next, the MIC-GCN module constructs the original graph and pseudo-graph. After ex-tracting local features through convolution, local and global features are aligned by maximizing mutual information. Finally, supervised contrastive learning is introduced to optimize the sentiment representation space and enhance the ability to distinguish sentiment polarity, with sentiment polarity judgment completed through pooling and classification layers. Experimental results demonstrate that on three public datasets—Twitter, Lap, and Rest—the accuracy of the proposed model is 0.1% percentage point, 1.71%, 0.3% higher than that of the best baseline models such as SSEGCN(Self-Supervised and Aspect-aware Graph Convolutional Network)、T-GCN(Type-aware Graph Convolutional Network) and DGEDT(Dependency Graph Enhanced Dual-Transformer network), respectively. The proposed model ef-fectively enhances semantic-syntactic coordination, improves the efficiency of capturing global features and the distinguishability of senti-ment polarity, demonstrating excellent performance in ABSA tasks.