Aspect mining in drugs reviews has focused on extracting relevant information such as adverse reactions, efficacy of a drug, symptoms and conditions of patients. In our work, a new unsupervised and knowledge-based method is proposed for extracting aspects in drug reviews. The proposed solution is based on linguistic features, more specifically dependency paths in the syntactic tree of a review. The quality of the dependency path rules was investigated in a number of experiments in review corpora associated to three different diseases. Promising results were achieved compared to previous work.
展开▼
