spaCy
spaCy is an open-source library for advanced natural language processing in Python and Cython, built for real products (“industrial-strength NLP”). The README calls it commercial open-source software under the MIT licence; the site footer credits Explosion.
Repository snapshot (GitHub API, 2026-11-19)
| Item | Detail |
|---|---|
| Repo | explosion/spaCy, MIT |
| Created | 2014-07-03 |
| Activity | last push 2026-10-27; latest release v3.8.18 (2026-10-26) |
| Size | 34,075 stars, 4,746 forks |
Capabilities
- Neural models for tagging, parsing, named entity recognition and text classification.
- Pretrained pipelines, including transformer-based pipelines (since v3.0) and a cheaper CPU-optimised one, plus multi-task learning with pretrained transformers such as BERT.
- A production-ready training system, model packaging and deployment, custom components, and information extraction at scale.
- Languages: the README says tokenisation and training for 70+ languages, the site says 75+ languages and 84 trained pipelines for 25 languages; the two sources disagree.
Where it fits
spaCy is the production-oriented choice; nltk is the teaching and research toolkit. Transformer models and training come from hugging-face and pytorch. For LLM-based extraction, compare RAG pipelines that parse text first.
Sources
- spacy.io, the spaCy README and GitHub API (2026-11-19)