Nuance taps into deep learning to improve Dragon speech recognition by 24 percent

Mark Hachman
17 August, 2016
View more articles fromthe author

With the trends in personal computing favouring software that gets to know its users, the newest version of Nuance’s Dragon voice dictation suite of software – Dragon 15, announced this week, is right on track.

Dragon 15—including Dragon Professional Individual (US$300), Dragon Professional Individual for Mac (US$300), Dragon Legal (US$500) and Nuance Dragon Anywhere (free to install; subscriptions run US$15/month or US$150/year) – is based on Nuance’s new machine-learning technology. The company claims that this technology has improved recognition accuracy by at least 24 percent, thanks to its algorithm that learns your distinctive speech patterns over time combined with an improved capability to pick out speech from a noisy room.

Lawyers may buy the expanded Legal version, which is trained using a legal vocabulary of more than 400 million words, according to the company. But the majority of Nuance’s customers will probably invest in the Dragon Professional Individual version, which is designed for a more general vocabulary.

According to Nuance, the Dragon software not only offers improvements in picking up and understanding your voice, but also features an improved transcription mode for working with recorded audio. Since Dragon works off of a ‘model’ for each speaker, it could theoretically improve if it ‘hears’ the same speaker in multiple recordings. It’s still going to struggle, however, in accurately transcribing a recording of multiple speakers holding a conversation, executives said. But a ‘batch mode’ can also transcribe multiple recordings quickly, they added.

All three apps can also be synced with the Dragon Anywhere mobile app (iOS and Android), allowing transcriptions to begin on the go.

Dragon Professional Individual version 15 for the PC and Mac will be available as a digital download on 1 September in the US, UK and Australia. Dragon Legal will only be available in US English, also on 1 September. (Physical copies ship on 14 September, according to Nuance’s website.)

Why this matters: Windows already has speech recognition built into it – not only can you use the ‘traditional’ engine inside the Speech menu within the Control Panel, but you can dictate reminders and short emails using Cortana. But it’s mediocre, and can’t handle longer documents. If you’re the type of person who writes clearly, but types slowly, however, dictation software may be for you.

Leave a Comment

Please keep your comments friendly on the topic.

Contact us