On April 29, Tencent cloud officially released its own voice recognition models in finance, audio and video fields. The newly released model not only greatly improves the recognition accuracy, but also increases the support for Cantonese and Korean. Later, it will gradually open the support for Shanghai dialect and other foreign languages such as Japanese, Thai and Indonesian.
Officially released the exclusive model of financial industry, leading the industry in word accuracy rate
It is understood that speech recognition is widely used in the financial industry, but in the actual scene, many users are using dialects to communicate with outbound calls and customer service robots. In addition, there are a large number of special sentence patterns and vocabulary in the financial field. At present, the general speech recognition model in the market is in the situation of inaccurate recognition.
Based on these pain points, Tencent cloud AI team and WeChat Zhizhi jointly created a financial industry exclusive speech recognition model, the introduction of this model can not only effectively solve the above problems, but also in the recognition accuracy has been greatly improved. At present, the model has been in the financial field of intelligent external call, intelligent customer service, telephone recording quality inspection and other scenes landing. By customer measurement, the accuracy rate in the industry in the leading level.
Take the lead in creating an exclusive voice recognition model for audio and video, with the accuracy increased by 10%
With the rise of Internet live broadcast wave, how to use intelligent voice technology to quickly identify users' audio and video content, carry out accurate recommendation and unhealthy content filtering has become the core competitiveness of each major live broadcast and content sharing platform in the increasingly fierce market competition. But because the audio and video background environment is complex and belongs to the half-field, it needs a lot of data accumulation to realize accurate identification.
With its own accumulation of data in the field of audio and video, Tencent cloud in the industry first launched the audio and video domain exclusive speech recognition model, has been used in a number of audio and video live broadcast platforms and e-commerce live broadcast platform, through customer measurement, recognition accuracy increased by nearly 10%.
The richness of languages is further improved, and Tencent cloud speech recognition is accelerated
In order to meet the needs of different customer groups, Tencent cloud speech recognition has continued to make efforts in terms of language richness this year. Combined with wechat intelligent listening, Tencent international business speech technology laboratory, Tencent people to Chinese translation and other artificial intelligence laboratories, on the basis of the original common languages, Tencent cloud speech recognition has opened the recognition ability of Korean and Cantonese. Later, it will open the recognition ability of Shanghai dialect, Japanese and Thai dialect Speech recognition ability of foreign languages such as Chinese and Indonesian. After long-term polishing training and effect optimization, Tencent cloud speech recognition languages have been widely used in business scenarios such as conference record transcribing, video subtitles, telephone recording and quality inspection.
In fact, Tencent cloud with years of hard work in the field of intelligent voice, has gained the recognition of many authorities. Tencent cloud became the only cloud manufacturer in china to Gartner the magic quadrant of cloud AI developer services in the first official Magic Quadrant for Cloud AI Developer Services》 study released this year.
Tencent cloud AI voice product director Zhou Chao said: