Interactive memory networks based on syntactic dependencies for aspect-level sentiment classification

In recent years, sentiment analysis has emerged as a key area of research in natural language processing, with particular emphasis on aspect-level sentiment classification. This subfield focuses on identifying and analyzing sentiments expressed toward specific aspects within a sentence. Current approaches often combine aspect and context representations to evaluate sentiment, yielding commendable results. However, these methods typically overlook the importance of syntactic dependencies and tend to extract coarse representations of aspects and context, leading to interference from irrelevant information. To address these challenges, this study proposes an interactive memory network that leverages syntactic dependency information. By employing a graph convolutional network, the model captures richer contextual and aspect representations. Additionally, an interactive memory network module is designed to enhance the precise feature relationships between aspects and context. To evaluate the model’s effectiveness, comprehensive experiments were conducted on four widely used benchmark datasets, demonstrating its superior performance in sentiment classification tasks.