Summary

If you're looking for a capable, easy-to-learn natural language processing toolkit, you can't go wrong with spaCy. They have an excellent tutorial that walks you through basic tasks including:

1. finding words, phrases, names and concepts;
2. large-scale text analytics;
3. creating complex, production NLP pipelines;
4. training custom neural networks.

It's our opinion that spaCy offers an excellent tradeoff between usability and capabilities. While its core architecture is based on somewhat older NLP approach of Word2Vec, it makes using this still very powerful NLP technique extremely easy. You can accomplish a lot with this approach, including custom named entity recognition, custom entity linking (e.g. determining if two documents or two nouns are linked by a verb), and more. If you want to use more recent advances in NLP technology, including transformer-based machine learning models, spaCy's developers have made great strides in incorporating these powerful techniques as well, though, if you want to use the most cutting edge models, you'll probably want to explore other models like HuggingFace's transformers models.