Microsoft has announced its major achievement in creating a technology that recognizes the words in a conversation. Now its speech recognition system is as accurate as human.
“We’ve reached human parity,” Microsoft’s chief speech scientist Xuedong Huang, said in a statement.
Microsoft claims that their Artificial Intelligence and Research division has reported a speech recognition system that makes the same or fewer errors than human transcriptionists. It had 5.9 percent error rate which beats a 6.3 percent record set last month. 5.9 percent error rate is lowest ever recorded in the industry and is equal to that of people who were asked to transcribe the same conversation.
To achieve these levels of accuracy, researchers have employed a deep neural network to store training sets to help system recognize patterns from human input.
This milestone is notable in the history of AI when we are trying to make computers as smart as humans. At least we achieved the same thing in speech recognition.
It is worth to note that this is a parity and not a perfection. It still shows 5.9 percent error rate, and humans do this as well. It misheard a word like “have” for “is” or “a” for “the” in similar way we human do sometimes.
We still have a long way to go in the field of artificial intelligence. These kinds of milestone encourage researchers to work harder.
Source: Microsoft Blog
