bjascob/LemmInflect

Combines dictionary lookup with neural networks to handle out-of-vocabulary words, drawing morphology rules from the NIH's SPECIALIST Lexicon and using unigram frequencies for disambiguation. Works as a standalone module or integrates directly into spaCy via custom token extension methods (`lemma()` and `inflect()`). Accepts Universal Dependencies or Penn Treebank tags to specify target word forms, achieving 95.6% accuracy on 119K inflected words with minimal dependencies (only NumPy).

275 stars and 94,489 monthly downloads. Used by 3 other packages. No commits in the last 6 months. Available on PyPI.