In aspect sentiment analysis tasks, when one or more aspect words appear in the text, different aspects correspond to different sentiment tendencies. Therefore, it is necessary to accurately identify the corresponding emotional tendencies of different aspects. Concerning the problem that the traditional neural network model cannot accurately construct the text aspect feature and sentiment feature information, an aspect sentiment analysis model based on the LSTM-GateCNN network is proposed. Textual contextual semantic information is modeled through the LSTM network fused with attention mechanism to obtain deep semantic information, combined with the gated convolutional neural network to simultaneously model the textual and emotional information, and finally, text sentiment classification is performed at the softmax function layer. Experiments on the AI Challenger Chinese dataset have verified the effectiveness of the model, which has a further improvement in accuracy compared with the previous method.