DrugSenti-Rx: Analyzing the Sentiment Landscape in Drug Reviews via Aspect Infused Graph Convolutional Networks

Sentiment analysis (SA) has become more important in drug evaluations because to its significance to healthcare professionals, pharmaceutical corporations, and regulators. This analytical approach assists in comprehending public viewpoints and attitudes towards drugs, therefore enabling the evaluation of the advantages and disadvantages of medications. Although neural network frameworks have made progress in aspect-level sentiment analysis, certain models face difficulties in pharmacological analysis because they fail to consider syntactic constraints and long-range word dependencies. This study introduces an innovative method for analyzing sentiment at the aspect level. The approach relies on a distinctive model known as Aspect-Infused Graph Convolutional Network (AIGCN). This work does a performance comparison between AIGCN and three others baseline models, namely LSTM, Attention model, and CNN-based model, using the UCI-ML drug reviews dataset acquired from Kaggle. The empirical results consistently demonstrate that AIGCN surpasses the baseline models in terms of both accuracy and F1 score. The outstanding performance of AIGCN can be ascribed to its effective integration of syntactic dependency information by employing graph convolutional networks (GCNs). The ability of AIGCN to efficiently capture word dependencies across vast distances enables the provision of dependable sentiment analysis.