Dr. Sridhar Dasaratha (Head of AI Research, Ernst & Young GDS)
Title
The Evolution of Document Intelligence
Abstract
Organizations often need to read, understand and interpret an ever-growing number of structured and unstructured business documents. In this talk I will start with an overview of document intelligence and discuss how methods for document analysis have evolved. While many deep learning methods are available to solve natural language processing tasks, in enterprise applications it is critical to combine domain knowledge and deep learning. I will introduce DeepPSL, a method that encodes heuristics and other domain knowledge using probabilistic soft logic with predicates that can be deep learning models. I will then go through examples on document classification and entity resolution demonstrating the effectiveness of DeepPSL.
Bio
Dr. Sridhar Dasaratha heads AI Research at Client Technology, EY Global Delivery Services. Sridhar has expertise in deep learning and natural language processing and has extensive experience in developing machine learning algorithms and solutions for multiple business applications. In his current role, he is responsible for evaluating, applying and enhancing emerging AI technology for tackling challenges in document intelligence. Presently his research is focused on key aspects of large language models, such as prompt engineering, retrieval augmented generation systems and fine tuning. Prior to joining EY, Sridhar led research teams to develop novel AI approaches and systems in diverse domains such as security, healthcare, renewable power and transportation. Sridhar is the author of 18 published or granted patents and has published 9 academic papers.