NVIDIA GTC 22 Session: Building a GPU-Powered Conversational AI Ecosystem for Medical Consultation & Contact Centre Decision Support

Conversational AI applications are comprised of a hierarchy of deep learning natural language processing (NLP) models and speech technologies. Making these many different components work together to realize robust real-world applications a complex, both in terms of model interaction and computing resource orchestration.

Deploying NLP models for processing speech requires end-to-end speech recognition, appropriate segmentation, and robust NLP model training processes that can cope with speech recognition transcription errors. In addition, effective conversational AI applications rely on numerous preprocessing models such as punctuation restoration, true casing, question classification, negation, and many others to ensure robust operation and reduction of false positives.

We present real-world contact center and medical consultation decision support cases to illustrate how such complex conversational AI ecosystems should be constructed.

Nigel Cannings, CTO at Intelligent Voice presented this session at NVIDIA GTC 2022.