This article originally appeared on our sister site, TabTimes.
Most of us probably use some form of speech recognition software at few times a day, but these bits of software are not always quite spot on at understanding what we say. Fortunately, Microsoft has just announced a breakthrough in speech recognition technology that transcribes spoken words as well as you and me.
A team at Microsoft Artificial Intelligence and Research have unveiled a new speech recognition system that has reached ‘human parity’ for the first time. In other words, the software is as accurate at identifying words as we humans are. Microsoft boasts of just a 5.9 percent error rate, which is roughly the same as professional transcriptionists achieved against the industry standard Switchboard speech recognition task.
Geoffrey Zweig, who manages the Speech & Dialog research group, attributed the accomplishment to the use of the latest neural network technology and a neural language model that groups similar works together for faster generalizations.
Of course, recognition is not the same as understanding, and Microsoft says that it plans to make use of the new system to improve the functionality available with its Cortana digital assistant, speech-to-text transcription software, and Xbox voice search tools. Although we don’t have any information on exactly how Microsoft’s new system will benefit these pieces of software just yet.
All of that being said, Microsoft researchers still say that they have a lot of work still to do to ensure that speech recognition works well in less ideal real world scenarios. Blocking out street traffic, picking out a speaker in a room full of conversation, and supporting a wider range of languages and accents are still works in progress. After all that, longer term research wants to teach computers not only to transcribe acoustic signals, but to properly understand language in the way that humans do. But that is still a long way away.
Virtual assistants have hit their stride this year, what with the new Google Assistant and Amazon’s Echo proving rather popular, so Microsoft’s speech improvements have come at just the right time for the company. It’s going to be interesting to see how Google Assistant, Alexa, Cortana, and Siri evolve differently over the coming years.