Shulou(Shulou.com)11/24 Report--

According to CTOnews.com news on December 22, Alibaba Dharma House today released a new generation of speech recognition model Paraformer, which is suitable for voice input method, intelligent customer service, car navigation, meeting minutes and other scenarios.

According to reports, this is the industry's first non-autoregressive end-to-end speech recognition model with a maximum reasoning efficiency of 10 times higher than the traditional model, and the recognition accuracy ranks first in many authoritative data sets. At present, the model is open to the whole society in the magic community.

The data show that with GPU reasoning, different versions of Paraformer can improve the reasoning efficiency by 5-10 times. At the same time, Paraformer uses a low frame rate modeling scheme with 6 times downsampling, which can greatly reduce the amount of computation and support the efficient reasoning of large models.

CTOnews.com learned that Ali Damo said that Paraformer is the next generation of "killer's mace" level speech recognition basic model developed by Alibaba, and will be widely used in conference minutes products such as "listening comprehension", voice-to-text nailing, Gao de navigation and other scenarios in the future.

Related readings:

"Ali Damo House launched the AI model community" Magic Dai ": open source 300 + AI model, you can experience Chinese painting AI"

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.