How U.S. Bank is Driving Rapid Advancements Across Business Units

, Head of AI Research, SVP, AICoE, U.S. Bank
, Executive Vice President, Chief Analytics Officer and Head of Artificial Intelligence, U.S. Bank
Developments in artificial intelligence – specifically in natural language processing, computer vision and graph analysis – are experiencing tremendous and rapid advancements, particularly in the technology and application areas they impact. The enterprise analytics group at U.S. Bank created a Center of Excellence to apply AI to provide solutions across the bank’s various business units. This COE embraces an open innovation model, where home-grown solutions for the bank’s problems are juxtaposed and contrasted with commercial offerings. We are creating ensemble solutions that are more performant, use case specific, extensible and – most importantly – explainable. In this presentation, we will walk through the development of the COE and highlight a specific use case (Information Extraction) that showcases the need for internal development, the sensitive nature of the data, ensemble development and scaling up with GPUs.
Information Extraction from financial documents (collateralized loan obligations, commercial loans, mortgages) requires not only natural language understanding of the original loan document, but also the amendments and notices sent periodically. Moreover, banks acting as a trustee of these instruments receive millions of loan notices per year. This effort focuses on agent bank notices entity extraction using state-of-art Natural Language Processing (NLP) methods. The objective is to extract information from loan notices on receipt and update the information automatically in a system of record. We have developed an ensemble of deep learning algorithms, using advanced transformer models, that extracts the required entities with human-level accuracies. This process can be scaled to process hundreds of thousands of loan notices in a single day using NVIDIA servers.
活动: GTC Digital September
日期: September 2022
级别: 商务 / 行政
话题: Data Science
行业: 金融服务
语言: 英语
话题: Conversational AI / NLP
所在地: