|This is the talk page for discussing improvements to the Dragon NaturallySpeaking article.|
|This article is of interest to the following WikiProjects:|
|This page was nominated for deletion on 12 March 2010. The result of the discussion was Withdrawn by nominator.|
One installation - one voice?
A friend told me that DNS "only recognizes one voice" so it is useless for transcribing interviews or such. Is this true? --03:36, 14 July 2009 (UTC)
- Yes that is correct it will not work with interviews. —Preceding unsigned comment added by 184.108.40.206 (talk) 23:21, 18 September 2010 (UTC)
- Not exactly true. You can set up multiple users on a single installation of DNS, so that, e.g., each member of a family can have their own voice and language model and be recognized appropriately without confusion, as long as they always log in to their own user profile within DNS. And DNS has for a long time had the ability to recognize from recorded speech, with correction afterward. But it will probably do quite poorly on a single recording with multiple speakers. Proper ASR with any ASR software would probably require an intermediate step of speaker identification to split the recording into single-speaker segments, not only recognizing "There is a change of speaker here", but also "The new speaker is speaker A" (who was the speaker in, e.g., segments 1, 3, 6, and 9), so that all of each speaker's segments can benefit from correction on each other. --Thnidu (talk) 18:46, 7 August 2012 (UTC) (Senior Linguist, Dragon Systems, 1990-2001
For transcription, it only works with one voice. Each copy of program, installed on a computer, is licensed for one user. However, you can have as many user profiles as you want, so many families create one profile for dad, one for mom, and so on. When you do live dictation, you choose the correct profile. Even with off line transcription, you have to choose the correct user profile, and again, it only works with that user only. — Preceding unsigned comment added by 220.127.116.11 (talk) 06:45, 3 August 2014 (UTC)
History needs major expansion
The history of Dragon software, technology, products, and applications needs a major expansion. This software has a unique and valuable history. Its role in the development of speech recognition and applications should not be lost. Usually when companies write Wikipedia articles they sound like marketing materials. I hope the current owners might put that aside and see how an accurate and full history would be viewed as a major boost of reputation. Completing this task, if it can ever be fully complete, might require calling on the former stakeholders and employees, maybe even competitors. Johnswolter (talk) 20:02, 3 January 2012 (UTC)
I just added a reference to a journal article in which James Baker proposed a system called DRAGON, which I'm sure influenced the commercial product that he came out with 7 years later (though I haven't found a citation yet that explicitly points to that model as the precursor to the commercial system). Showeropera (talk) 17:26, 18 December 2012 (UTC)
- Certainly the name was precursor. The software may also have been ancestral, but I'm not sure of that. And in any case, my memories of company literature from my ten or so years there as Senior Linguist are inadmissible as references (WP:OR). But maybe someone will dig those materials up someday.
- But wrt User:Johnswolter's paragraph, Wikipedia doesn't allow companies to write or edit articles about themselves, I'm pretty sure. --Thnidu (talk) 05:49, 19 November 2014 (UTC)
This topic title is deeply correct. James K. Baker singlehandedly invented modern speech recognition as a doctoral student at Carnegie-Mellon University; his (then an unpatentable "method") invention received the name "hierarchical Markov model". Baker did not just write a paper, as this article now says--he designed and built an entire prototype system at CMU. It was his Ph.D. dissertation project. He named it the "Dragon System", and, when tested alongside several other systems built over five years by multiple contractors with $50 million of (D)ARPA money, Baker's Dragon System beat all of them! This all was documented fairly at some length by noted MIT professor Dennis Klatt in a paper in the Journal of the Acoustical Society of America (JASA) in about 1976-78. Klatt's report covered the "ARPA Speech Understanding Research Project", "ARPA SUR" for short, which ran in parallel (1971-1976) with the much more famous ARPA project that gave us the ARPANET, which became the Internet.
After he received his Ph.D., Baker went with his wife, Janet (also a speech recognition person) to IBM, which had been following ARPA SUR development closely. They were not there long but IBM learned Baker's technology. The Bakers then went back to the Boston area and started their "Dragon Systems" company on the proverbial kitchen table. Versions of their company website before the Lernout & Hauspie disaster told the story of their early company days and growth. There was also a writeup in a business journal, possibly "INC." The entire story of the Bakers deserves a book-length treatment or even a movie. I met them during the ARPA SUR years when they visited the company for which I then worked but have not spoken to either of them since then. They are decent, good folks.LM6407 (talk) 06:35, 18 June 2015 (UTC)
The latest revision was the addition of a link
- http://dragonvoicesoftware.com Dragon Voice Software
under "References". If it had any business being on the page, it should be under External Links. But in fact a whois search reveals that this domain is not owned by the manufacturer, Nuance Communications, but by a private individual, most likely a retailer. So I took it out. --Thnidu (talk) 22:59, 20 June 2014 (UTC)
64-bit or 32 bit
I think it should be mentioned in the article that NaturallySpeaking is still (as of version 13) a native 32-bit application with code added to make it run on 64-bit systems. — Preceding unsigned comment added by 2601:6:2180:D0:3C7C:EEED:EDF0:4EF8 (talk) 11:08, 23 March 2015 (UTC)