On 30 June 2010 16:43, silner <silnerwil...@gmail.com> wrote: > On Tue, 29 Jun 2010 22:31:32 +0100, Roy Jamison wrote: > >> I have found Dragon NS to be quite surprisingly accurate after 10-15 >> minutes of training, so I don't think the perfect model is too far away, >> at least for closed-source payware. > > Someone told me, a while a go I must admit, that Dragon NS worked well > for women and men with high voices, but not nearly so well with deep > voices. I was told it was so bad as to be unusable for deep voiced users?
I wrote a whole series of reviews for PC Magazine UK in the late 1990s using IBM ViaVoice, after lacerating the palm of my right hand in a nasty washing-up accident. It worked amazingly well - with training it could distinguish between "a gateway on the network" and "a Gateway PC" just from the slight emphasis on the proper noun. I /do/ have a pretty deep, booming voice, too. It's not true, I think. Sounds like FUD to me. Speech recognition was working well a decade ago, on for-the-time poxy hardware. It needs training, though, and it does have a small error rate - 2-3%. This is enough to be irritating if you can type reasonably well. What we /don't/ yet have is speaker-independent continuous speech recognition, i.e., without training. That's hard. But speaker-dependent stuff, if you take the time to train it, is fine. It's just that typing is quicker. -- Liam Proven • Profile: http://www.linkedin.com/in/liamproven Email: lpro...@cix.co.uk • GMail/GoogleTalk/Orkut: lpro...@gmail.com Tel: +44 20-8685-0498 • Cell: +44 7939-087884 • Fax: + 44 870-9151419 AOL/AIM/iChat/Yahoo/Skype: liamproven • LiveJournal/Twitter: lproven MSN: lpro...@hotmail.com • ICQ: 73187508 -- ubuntu-uk@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-uk https://wiki.ubuntu.com/UKTeam/
