What is Flair library?
Flair library was developed by Zalando Reserch and it’s a simple NLP library built using PyTorch. Flair has few pretrained models for different NLP tasks such as:
Training custom models
Flair is currently the SOTA library for NER and PoS NLP tasks!
What’s so special about Flair?
Flair contains all the popular SOTA word embeddings such as GloVe, BERT, ELMO, and character embeddings. It also allows you to combine different word embeddings together. Most importantly, Flair provides the ‘Flair Embedding’ which it’s contextual string embeddings.
What is the Flair Embedding?
Contextual string embeddings takes advantage of the internal principles of a trained character language model to produce a novel type of word embeddings. There are two primary factors:
The words are trained as characters, similar to character embeddings
The embeddings are contextualised by their surrounding text, meaning that the same word could have different embeddings based on the context