Find out more about the used in Estonian public broadcasting?

Essential for converting spoken words into readable text (e.g., changing spoken numbers into digits).

High-speed, chaotic, or overlapping dialogue (common in talk shows) still results in higher word error rates (13.4%) compared to news.

Automated systems in Estonia utilize speech-to-text engines based on Kaldi-based TDNN-F models , which are specifically trained to recognize Estonian speech segments, punctuation, and normalize text.

Estonian Public Television (ERR) utilizes these systems for live broadcasts, and the Estonian Parliament uses them for live video feeds of parliamentary sessions. Performance Metrics (Quality Analysis)

Automated insertion of periods, commas, and question marks to improve readability, crucial for accurate comprehension of live news. Limitations