For recommender systems to personalize suggestions by understanding nuanced user preferences toward specific product features or service attributes, fine-grained sentiment signals are critical. The main purpose of Aspect-Based Sentiment Analysis (ABSA) is to identify the sentiment polarity of a given aspect in a sentence, thereby providing these essential signals. Recent investigations have demonstrated the excellent performance of incorporating dependency parsing trees with graph convolutional neural networks (GCNs). However, these GCN-based models mostly employ aspect representations for final sentiment classification, lacking exploration of global information. Moreover, the performance of GCN-based models strongly relies on the quality of the dependency parsing tree and may yield suboptimal aspect representations. To address these deficiencies, we propose a graph convolutional network with semantic and syntactic knowledge (SSK-GCN). Specifically, a gating mechanism is first employed at each GCN layer to consider the given aspects when aggregating information, generating aspect-oriented representations. Then, a self-attention mechanism is utilized to complement the dependency parsing tree by enriching the syntactic structure, thereby generating global-oriented representations. Finally, aspect-aware attention is used to measure the sentiment relevance between optimal aspect representations and global-oriented representations for final sentiment prediction. Experimental results on five benchmark datasets demonstrate that our SSK-GCN outperforms other state-of-the-art models in both accuracy and macro-F1 metrics.