pemistahl/lingua-rs

Based on the README, here's a technical summary: Combines rule-based and statistical Naive Bayes classification without neural networks or external dictionaries, enabling offline language detection from single words to full sentences across 75 languages. Trained on Leipzig University corpora with separate train/test splits from news data, delivering measurably higher accuracy on short text than competing libraries like CLD2 and Whatlang. Includes minimal configuration requirements and ships with bundled language models for immediate use without API dependencies.

1,067 stars and 235,303 monthly downloads. Actively maintained with 2 commits in the last 30 days.