Aspect-Level Drug Reviews Sentiment Analysis Based on Double BiGRU and Knowledge Transfer

In 2019, researchers Yue Han, Meiling Liu, and Weipeng Jing developed a model called PM-DBiGRU for aspect-level sentiment analysis in drug reviews. The model used pretrained weight from short text-level drug review sentiment classification and two BiGRU networks to generate bidirectional semantic representations of the target and drug review. An attention mechanism was used to obtain target-specific representation for aspect-level drug review. The researchers also introduced a dataset, SentiDrugs, for aspect-level drug review sentiment classification. The PM-DBiGRU model outperformed other models, with its accuracy and Macro-F1 being 1.47% and 2.39% higher than those of the PRET+MULT model, respectively. However, the model had limitations in cases where there was no direct sentiment expression towards the target, certain background knowledge was required to judge the sentiment polarity, or the sentence structure and sentiment expression were complex.