Compliance Program Enhancements using AI: Natural Language Processing with LEX
LEX is a Natural Language Processing Platform which was built by the Technology Group and Supervision. It is a language extraction tool that identifies language of interest. Its primary use case has been to identify language within Supervision’s continuous monitoring documents but it can be applied to many other use cases. LEX is the first custom-built NLP software put into production in FRBNY and has since been rolled out to 800 users within NY Supervision and the system. This talk will cover:
• Business Problem: BSA/AML
• Application of AI/ML/NLP • Language models and how word embedding changed everything!
• How a language model trained on news articles is good, but one trained on relevant documents is so much more powerful, particularly when blended with the news model.
• The importance of scheduled retraining and making sure our models have an up-to-date vocabulary
Harry Mendell specializes in artificial intelligence, machine learning, natural language processing and data architecture. He applies these techniques to innovative solutions for the Federal Reserve. He is in the Data Architecture Group and is also co-chairman of the AI Innovation Roundtable which helps to promote the use of AI and machine learning throughout the system. Harry has a BS and MS in Computer Science from the University of Pennsylvania. His thesis was in artificial intelligence and computer vision. He then worked at Bell Labs with the original Unix team to design the first Unix based workstation. He then joined the financial sector focusing first on the pricing and trading of equity and credit derivatives and later risk management and compliance. The re-emergence of AI and NLP around 2012 led Harry back to his original interests in these areas and he formed a startup to research and develop applications in FinTech which included financial compliance tools. Realizing that this mission was too large and too important for a startup he decided to take a more direct route to introducing these ideas and is now working at the Federal Reserve.