Deep Learning – Based Models for Drug-Drug Interactions Extraction in the Current Biomedical Literature

Drug-Drug interactions (DDI) constitute important and useful information for medical staff and patients, since it provides information on the effects of medications co-administered to a patient produce during therapy. DDI are scattered in a vast collection of pharmaceutical documents, which is why many approaches have been applied in extraction tasks, such as support vector machine (SVM), recurrent neural networks (RNN) and long-short term memory (LSTM). This article presents the background and a systematic literature review related to the deep learning models of extraction of DDIs found in the current biomedical literature, and summarizes the existing evidence on the most prominent approaches in that field; alike, it is identified the used methods and the pending challenges. The possibility of investigating the creation of an alternative model with better performance, opens up.