Aspect-Level Sentiment Analysis Based on Syntactic Mask and Graph Convolutional Network

In this age of social media and online review platforms, the significance of aspect-level sentiment analysis, which seeks to determine sentimental trends regarding specific aspects within text, has grown substantially. Traditional sentiment analysis techniques frequently overlook syntactic structure information, which is crucial for accurately interpreting the emotional nuances of sentences. In order to productively improve this defect, our research forms an innovative neural network based on the Syntactic Masked Graph Convolutional Network (SM-GCN). This network model leverages syntactic information and the capabilities of graph convolutional networks (GCN) to get the weight of each aspect of a sentence and its intricate relationship to the overall text. By incorporating syntactic mask matrices, the model hones in on key segments that are pivotal for sentiment analysis, thereby enhancing the accuracy of sentiment classification. The model uses the abstract graph structure principle of graph convolution to integrate the information of the next node by using the transfer matrix of the fusion syntactic mask matrix to obtain the enhanced performance of the model structure. The SM-GCN model was tested on multiple public ABSA datasets, and the findings indicate that the innovations we proposed yield notable outcomes in a series of analytical tasks that we have given at the emotional level, and have significant advantages in accuracy and F1 score, which proves the adaptability and practical value of the model.