Alibaba Cloud makes its high-accuracy speech recognition model (96.04%), Deep Feedforward Sequential Memory Network (DFSMN), open source on GitHub.