On April 12, China Mobile held a “5g new call, foresee the new future” product conference at the speed skating Hall of Shougang Park National Winter Training Center, officially released 5g new call products, and announced that some terminals have supported 5g vonr ultra clear call business throughout the country to create a new media with visualization, multi-media, high perception and full interaction.
will cover all new models
At the press conference, Shou Jianguo, general manager of China mobile market operation Department, introduced that compared with voice and video calls on the Internet, 5g new calls can realize high-definition video calls based on China Mobile 5g network, with stable call quality, almost no delay and uninterrupted incoming calls. At the same time, combined with AI technology, 5g new call can realize the functions of Chinese and English real-time translation, voice to text and so on. In addition, it also provides screen sharing, remote cooperation and other characteristic functions. Shoujianguo also said that at present, 5g new call functions are being adapted in major mobile terminals. By July this year, all new models of terminals will support 5g new calls.
From the function of 5g new call, it is expected to become a “5g killer application”. Behind this business widely favored by the industry, 5g network is the core foundation, media interaction is an important function expansion, and terminal support is the guarantee. Another key is AI intelligent voice technology.
not only hear, but also understand
5g new call enables both sides of the call to realize voice to text, and can also be translated in real time on the screen, so that the caller can “understand” the content of the call while listening. Seemingly well understood technology, there are a lot of black technologies behind, such as speech recognition, oral comprehension, voice simultaneous interpreting and so on. The accumulation of non deep AI voice technology is not to be done.
The black technology behind these mainly comes from the head enterprise Iflytek Co.Ltd(002230) of intelligent voice and artificial intelligence. Therefore, Iflytek Co.Ltd(002230) has also become the official partner of China Mobile’s 5g new call.
It is reported that Iflytek Co.Ltd(002230) is the exclusive supplier of official automatic voice conversion and translation for Beijing 2022 Winter Olympic Games and winter Paralympic Games. Statistics show that the multilingual voice and language service platform for the Winter Olympics scene adopts Iflytek Co.Ltd(002230) ‘s “automatic voice conversion and translation” technology, and supports 60 language speech synthesis, 69 language speech recognition, 168 language machine translation and 6 language interactive understanding. Among them, the accuracy of key language translation has reached 95%, and the average response time of each sentence translation is no more than 0.5 seconds.
oral experience is smoother
There are a large number of colloquial expressions in the call scene. The colloquial expression is different from the standard written text. Generally, the content expression does not meet the grammatical norms, the modal particles are too heavy, and the repetition is redundant, which makes the literal translation appear obvious “machine translation” traces, and puts forward higher requirements for the application of machine translation in the call scene.
According to the technical director of Iflytek Co.Ltd(002230) , three measures have been taken to optimize the oral scene of 5g new calls: first, human-computer cooperation marks common oral data and supplements oral bilingual training; Second, the unsupervised / weakly supervised training method is systematically used. Based on a large number of monolingual data of the source and target language of oral expression, the self training and back translation algorithms are used to strengthen the translation model and language model, so as to realize the enhanced training of the characteristics of oral expression; Thirdly, for the post-processing stage of speech recognition, the modules of smooth and regular modal particles are designed to make the spoken expression written as much as possible, so as to reduce the “traces” of machine turning and help users better understand.
The AI technology behind China Mobile’s 5g new call comes from the profound accumulation of adhering to the independent innovation of source technology in the past Iflytek Co.Ltd(002230) 23 years Iflytek Co.Ltd(002230) is an international leader in artificial intelligence speech recognition, machine translation, semantic understanding and other technical fields, and has won the championship in many international technical competitions.
