Search
Alibaba Tongyi Lab Unveils Speech Recognition Model: Fun-ASR1.5
10 hour ago /
Read about 0 minute
Author：小编
On April 20, 2026, Alibaba Tongyi Laboratory proudly announced the launch of its innovative speech recognition model, Fun-ASR1.5. Built upon a unified architecture, this versatile model boasts the capability to handle 30 distinct languages, seven major Chinese dialects, and more than 20 regional accents, ensuring precise transcription even for ancient poetry recitations. In real-world dialect scenarios, the model has demonstrated a remarkable 56.2% reduction in character error rates, achieving over 90% accuracy for five dialects and surpassing 80% accuracy for fifteen dialects. Presently, the model is accessible on Alibaba Cloud's BaiLian platform, providing API services for seamless integration.
-
C114 Communication Network -
Communication Home
7 X 24 Track global technological trends
