After clicking “Watch Now” you will be prompted to login or join.
Building a smart language understanding system for conversational AI with Hugging Face transformers
Lysandre Debut, Hugging Face Inc
GTC 2020
In this session, HuggingFace showcases an example of a natural language understanding pipeline to create an understanding of sentences, which can then be used to craft a simple rule-based system for conversation. They'll leverage the famous HuggingFace transformers and showcase the powerful yet customizable methods to implement tasks such as sequence classification, named-entity recognition, natural language generation, or question answering. These tasks will be joined to create a basic NLU pipeline to get the most out of a sentence or text, using transformer models such as BERT to provide state-of-the-art results. These methods leverage the PyTorch or TensorFlow numerical computation frameworks, which can leverage the power of GPUs to radically speed up the inference.