2023 HPCC Systems Community Summit: Building Trustworthy and Auditable Digital Human Readers...

Подробнее о видео

Building Trustworthy and Auditable Digital Human Readers Superior to Humans with NLP++ - David de Hilster, LexisNexis Risk Solutions & Ashton Williamson, Clemson University

Machine Learning, Neural Networks and Large Language Models are statistical and opaque, human memory is fallible, and NLP++ is now stepping in to fix these problems and build trustworthy systems in areas such as law enforcement, healthcare, and sentiment analysis. Not only can NLP++ perform better than humans, but it can build systems that eventually can be better than human experts. Unlike the statistical methods of ML, NN, and LLM which are opaque and not auditable, NLP++ is a verifiable and auditable technology which is a game-changer in today’s critical systems. NLP++ is a glass box computer programming language and framework that seamlessly integrates with HPCC systems to provide sophisticated text processing that until now, have been only achievable by humans. We will look at work being done at Clemson to help solve the medical coding task that is currently done by human readers and has a 50% failure rate.

© 2023 LexisNexis Risk Solutions