Document Classification and Information Extraction framework for Insurance Applications

Daniel Dickinson, Ananth Raj GV, Glenn Fung, Qian You, Eric Bunch

Abstract: Document Intelligence is an essential subclass in the field of machine learning. It plays a vital role in insurance applications and other sectors. In this work, we showcase a business application that uses two different but Complimentary techniques: document classification and entity extraction. We also provide an overview of an end-to-end production level system that incorporates deep learning models deployed at scale. The system’s backbone relies on trained models carefully analyzed and designed to generalize well on existing and future usecases. Through empirical evidence, we provide insights into several models trained on our insurance-related datasets and highlight models that have shown good performance across multiple datasets in our real-world insurance setting.