Short text classiﬁcation is a fundamental problem in natural language processing, social network analysis, and e-commerce. Traditional approaches for classifying text do not generalize to short texts, due to the lack of structure that is prevalent in longer sentences and paragraphs. More recently, deep learning-based methods have been applied to this problem, with limited success. To overcome the limitations of textual features in shorttext, in this paper, we propose to apply Graph Convolutional Network(GCN) to classify shorttext. The method uses side-information present in the dataset to construct a graph and then learn a Short Text Graph Convolutional Networks (ST-GCN). We demonstrate the efﬁcacy of our method in the context of two variants of the problem in the e-commerce domain where the text is short: product query classiﬁcation and product title classiﬁcation. The use of a GCN to represent our corpus allows us to capture dependencies between the text samples and this additional construction allows us to surpass several baseline methods. Our model achieves state-of-the-art results on both proprietary and external datasets, where ST-GCN outperforms other methods by up to 5.89 % in classiﬁcation accuracy. Furthermore, we show that compared to baseline methods, ST-GCN is relatively more robust to noise in textual features.