China has released its first large-scale speech model that supports the mixing of 30 dialects

Chinese operators have released the first large-scale voice model that supports 30 dialects.

Comprehensive China News Service and Securities Times reported that China Telecom artificial Intelligence Research on Saturday (May 25) at the seventh Digital China Construction Summit, released the industry's first voice recognition model that supports the free mixing of 30 dialects - Star Super multi-dialect speech recognition model, breaking the dilemma that a single model can only identify a specific single dialect.

The large model can recognize and understand more than 30 dialects such as Cantonese, Shanghai dialect, Sichuan dialect, Wenzhou dialect, and so on, which is the large speech recognition model supporting the most dialects in China.

The Star Speech large model is also the industry's first open source speech recognition large model based on discrete speech representation, which reduces the bit rate of speech transmission during reasoning by tens of times through a new modeling paradigm of "from speech to token to text".

At present, the star voice large model has been applied in Fujian, Jiangxi, Guangxi, Beijing, Inner Mongolia and other places of China Telecom intelligent customer service pilot. After accessing the Star Grand model, the intelligent customer service can understand 30 dialects in a second and handle about 2 million calls per day.