Predicting the activity and properties of chemicals based on RoBERTa

Machine learning allows us to solve problems by learning big data. In recent years, machine learning has been widely used in various fields, including solving various problems in chemistry. Examples include in computational chemistry, in the discovery of new drugs and even in the prediction of the synthesis of substances. In this work, we propose BPE and RoBERT, the latest technologies in natural language processing, to predict the activity and properties of chemicals. The results obtained were compared with the results given by the ChemBERTa model.