Tirilish

Tirilish is a multi-page platform for Uzbek language infrastructure: collecting a corpus, governing a lexicon, and giving contributors a place to work on both together.

Underneath sits an Uzbek text engine, packaged on its own as @tirilish/uzbek, so the website and any other tool run exactly the same code:

• Normalizer: repairs conservative mojibake, normalizes Uzbek apostrophes, cleans spacing and punctuation, and detects the script.
• Tokenizer: offset-preserving word, number, punctuation, symbol and URL tokens, sentence spans, and mixed-script warnings. It also recognizes emails, hashtags, mentions, abbreviations, dates, times, legal references and Roman numerals.