spaCy models for texts of the 16th to 20th centuries. Both do part-of-speech tagging (coarse pos_ and fine-grained tag_) and lemmatization (lemma_).
| Model | Language | Training data | Download |
|---|---|---|---|
| historic-french-model | French | 50,000 sentences from ARTFL-Frantext | fr-v7.0 |
| historic-english-model | British English | 65,000 sentences from EEBO-TCP, ECCO-TCP and Project Gutenberg | en-v2.0 |
Each model's README gives its scores, installation and rebuilding steps.
fr/,en/: one directory per model, with its training data, gold sets, tokenizer, annotation guidelines and scripts. Run a model's scripts from its directory.shared/package_model.py: makes a trained pipeline load without this repository's code.
Release tags start with the model's language (fr-v7.0, en-v1.0). The tags without a prefix, v1.0.0 to v6.0, are French releases from before the English model was added.
This repository was previously named spacy-historic-french-model; old links redirect here.