China Telecom has released its first large-scale voice model supporting 30 dialects...
【数据猿导读】 China Telecom has released its first large-scale voice model supporting 30 dialects
On May 26, the China Telecom Artificial Intelligence Research Institute (TeleAI) released the industry's first large speech recognition model that supports the free mixing of 30 dialects, Star super multi-dialect speech recognition large model, breaking the dilemma that a single model can only identify a specific single dialect, and can simultaneously identify and understand more than 30 dialects such as Cantonese, Shanghai, Sichuan and Wenzhou. It is understood that China Telecom Artificial Intelligence Research Institute has built more than 30 kinds of high-quality dialect databases, more than 300,000 hours, the first "distillation + expansion" joint training algorithm, to solve the problem of pre-training collapse under super-large-scale multi-scene data sets and large-scale parameter conditions, to achieve 1B parameter 80-layer model stable training.
来源：DIYuan
刷新相关文章
我要评论
不容错过的资讯
大家都在搜
