WordPredictor 1.4.0 - Major Corpus Expansion
What's New
Bundled n-gram corpus expanded from 5,096 bigrams and 32,625 trigrams (1.6 MB) to 45,859 bigrams and 300,384 trigrams (25.4 MB), trained on 11 million tokens from 116 Project Gutenberg books.
Predictions should be significantly more accurate with broader vocabulary coverage.
Details
- Bigrams: 5,096 → 45,859 contexts (754,990 entries)
- Trigrams: 32,625 → 300,384 contexts (1,043,510 entries)
- Data file: 1.6 MB → 25.4 MB (7.4 MB compressed in add-on)
- Load time: ~0.6 seconds (one-time at NVDA startup)
- Prediction speed: 5.7ms per prediction (well under 200ms threshold)
No code changes. The Kneser-Ney model, prediction logic, settings, key bindings, terminal detection, and all features are unchanged. This is a data-only upgrade. Real-time learning still works on top of the expanded corpus.
Install
Download the .nvda-addon file below and open it to install in NVDA.