Title: Building Financial Language Models: Data, Design and Benchmarks
Abstract: Domain specific language models are essential for gaining an in-depth understanding of financial documents. The development of such models requires curating data that reflects the distinct features of finance, incorporating specialized training strategies to exploit underlying characteristics and evaluating performance using realistic benchmarks. This talk will highlight our contributions to these three areas, focusing on the challenges related to financial terminologies, quantitative reasoning, relational structure and visual inference.
Bio: Natraj Raman is an AI Research Director at JPMorgan Chase, London. Prior to this, he worked as a Lead Data Scientist at S&P Global and as a Senior Research Scientist at Thomson Reuters. Natraj received his PhD in Computer Vision from the University of London and has more than two decades of industry experience in developing solutions for financial professionals. His primary research area is statistical machine learning with a particular focus on pattern recognition, information retrieval and representation learning. [more information]