On September 23, iFLYTEK officially released the latest generation speech recognition model, Spark-ASR-2.0. According to reports, through key technological innovations such as non-autoregressive and LLM enhanced autoregressive collaboration, joint enhancement of Chinese-English mixed text and acoustics, and dynamic context injection, Spark-ASR-2.0 has greatly improved the overall speech recognition effect. In particular, improvements are obvious in general recognition, complex acoustic scene recognition, context recognition, and smooth and standardized text.

Zhitongcaijing · 3d ago
On September 23, iFLYTEK officially released the latest generation speech recognition model, Spark-ASR-2.0. According to reports, through key technological innovations such as non-autoregressive and LLM enhanced autoregressive collaboration, joint enhancement of Chinese-English mixed text and acoustics, and dynamic context injection, Spark-ASR-2.0 has greatly improved the overall speech recognition effect. In particular, improvements are obvious in general recognition, complex acoustic scene recognition, context recognition, and smooth and standardized text.