Literature Classification of Individual Reports of Adverse Drug Reactions Based on BERT and CNN

In clinical practice, deaths due to adverse drug reactions and hospitalization and outpatient costs due to improper use of medication have risen sharply, becoming the major problems facing the safe and rational use of medication in the clinic. At present, retrospective analysis of adverse drug reactions and literature analysis are mostly based on publicly published literature, and academic literature, as one of the important data sources, is particularly important for how to process data automatically and in batch. In view of the unique expression of pharmaceutical text, text classification technology comparison experiments are conducted based on BERT and its combination model, to establish an efficient and fast classification method for the literature data of adverse drug reaction reports, so as to distinguish the types of adverse drug reactions and effectively warn the drug damage events. The experimental results show that the classification accuracy of the BERT model reaches 99.75%, which can accurately and efficiently classify the literature of adverse drug reaction cases, and is of great value and significance for assisting medical treatment and constructing structured data of medical text, so as to safeguard public health.