This article uses the main ideas of the "READ: Recursive Autoencoders for Document Layout Generation" paper and discusses its implementation in frontend scenarios.