Microsoft creates speech recognition system with human-level accuracy
In a remarkable leap forward a group of analysts and designers at Microsoft Artificial Intelligence and Research reported that they have made an innovation that can perceive words from a discussion pretty much and in addition a normal human. The group included that that the discourse acknowledgment framework makes an indistinguishable number of blunders from a human transcriptionist. "We've achieved human equality. This is a noteworthy accomplishment," Xuedong Huang, Microsoft's central discourse researcher expressed in a blog entry. As per a paper distributed on Monday, October 17 the specialists reported a word blunder rate (WER) of 5.9 percent against 6.3 percent reported a month ago. The specialists tried the discourse acknowledgment framework on the "Switchboard" discourse acknowledgment framework. Switchboard is an accumulation of recorded telephone discussions in English, Spanish, and Mandarin, initially discharged by the National In...