The software is based on Multilayered Attention Modeling (SMLTA) which has increased the accuracy of on-line voice authentication by 15%. Current input software is mainly used for mobile terminals, but in the future it will be applied to interaction between various smart devices and users. With the progress of voice input technology that mixes Chinese, English and dialect, users will realize “free talk”, even if Chinese and English are mixed, or if Mandarin is changed to home dialects. Baidu’s overseas input software also allows for Japanese identification, Hindi recognition, and has an Hindi-English mixed input function.
Japan JST China news, January 18, 2019
Baidu presents AI-based language recognition software which responds to dialects