Microsoft developed update of speech recognition “Cortana” like a human
After a long period, since 20 years of research Microsoft. Artificial intelligence team achieved a high level of success in Speech Recognition “Cortana”. The system they developed this time is good as a human being.
What’s new in Microsoft speech recognition?
According to the reports of TechCrunch, “word error rate” in this system has scaled to 5.9%from 6.3%. This is the same level as a professional human transcriber. In simple words, word error rate is the percentage of times of conversation. Even it is assured that speech recognition “Cortana” is enough good for conversations. Cortana has an accessibility feature, say, for deaf people.
This thing can understand those users who are using window10 on their laptop or PC. Overall, Microsoft speech recognition has a number of substitution rate and a higher number of deletion rate. The higher rate deletion rate can reflect a human bias to avoid outputting information. Even it can deal the productivity demands on a professional transcriber.
Last update of Microsoft speech recognition:-
Like Microsoft notes in paper and humans and other automated systems mostly fumble. Like the same sounds like “uh-huh” and “uh”. Microsoft confused by “uh-huh” which is a verbal nod for someone to speak ahead. And “uh” is like hesitation while speaking.
New update of Microsoft speech recognition:-
In this new update, there Is improved automated speech recognition “Cortana” system could be used in speech-to-text transcription service and enhance. Cortana accessibility features, say, for deaf people.
To improve these features Microsoft used 2,000 hours of training data to improve its neural network to recognize and complete the task. To parallelizing the data with it’s AI computational network toolkit on a Linux-based multi-GPU server farm.
Microsoft given statement that “there is milestone away to achieve well speech recognition “Cortana” which will work with a lot of noise all around”.