flair

Why we included this project

Flair is a Python NLP library built around a simple promise: load a pre-trained model and you can tag entities, score sentiment, or assign part-of-speech tags in a few lines of code, with no inference pipeline to hand-wire. It doubles as an embedding toolkit, letting you combine Flair's contextual embeddings with transformer-based ones and feed the result into your own models. Teams that want to train custom sequence taggers or text classifiers instead of calling a hosted API will find the PyTorch foundation makes fine-tuning on their own labeled data straightforward. Built-in biomedical support and models for many languages also widen the range of text domains you can take on without starting from scratch.