Advances in natural language processing (NLP) over the last year have been fundamentally changing the way we work with text-based data. Amazing libraries such as Transformers (by Hugging Face) and Flair (by Zalando Research) have made it easier to apply the most cutting edge models to generate insights from text data. However, we have found that some users of these libraries still had trouble when adapting these libraries to work with their data.
To address this challenge, Novetta developed AdaptNLP, an open source Python framework for applying and retraining NLP models. This framework enables users to apply state-of-the-art pre-trained language models, which can be fine-tuned for text classification, question answering (QA), entity extraction, and part-of-speech tagging. This tutorial will demonstrate how to use AdaptNLP to train a text classifier, train a simple QA model, and combine QA and named entity recognition (NER) to develop a timeline of events from unstructured text.