Home | Gallery | Forum | Services | Webmail | Archive | Links | Contact Us | About Us
OS2 World.Com Forum
OS2 World.Com Online Discussion Forum.
Index / OS/2 - General / Applications
author message
IBM Releases Speech Technology To Open Source; Chance to Port to OS/2-eCS?
Post a new topic Reply to this Topic Printable Version of this Topic Forward this Topic to your Friend Topic Commands (for administrator or moderators only)
BigWarpGuy
Premium member
in staff

View this member's profileSearch all posts from this memberSend an email to this memberhttp://home.comcast.net/~tomleem
posts: 2298
since: 12 Jan, 2001
1. IBM Releases Speech Technology To Open Source; Chance to Port to OS/2-eCS?
Reply to this topic with quote Modify your message
"IBM on Monday said it has handed over speech technology to two open-source groups to boost development of speech-recognition applications, which proponents say will eventually become commonplace within customer-service call centers, automobiles and elsewhere. ..."

Someone could port to OS/2-eCS?


BigWarpGuy
- - - - - - - - - - - - - - - - - -
OS/2 Warp-ed/eComStation-ed to the very end.
- - - - - - - - - - - - - - - - -
Date: 15 Sep, 2004 on 14:36
warpcafe
Premium member
in user

View this member's profileSearch all posts from this memberSend an email to this member
posts: 350
since: 26 Nov, 2002
2. Re:IBM Releases Speech Technology To Open Source; Chance to Port to OS/2-eCS?
Reply to this topic with quote Modify your message
last updated at 16 Sep, 2004 10:22 (1 times)
Just to clear up on one or two things:

People might say "why porting? We have VoiceType for OS/2 integrated."
...but:
VoiceType is somehow "older" technology requiring the "speaker" to use short pausings for the system to recognize well. In contrary, what IBM is somehow giving away is "ViaVoice" - a newer technology that was available for Windoze only and uses "continuous speech" recognition - thus "real" speech.
I would also appreciate if someone could investigate on porting possibilities at least, but I'm afraid that there is not much need in the OS/2 and eCS community: From what I have found, I am the only one to use speech recognition from time to time in OS/2. I never saw someone else use VoiceType on Warp even at warpstocks (europe). Maybe this could change with continuous speech recognition and a quite documented API for integrating speech recognition into applications for developers...
Maybe we should wait for a Linux port and then port it to OS/2?

Greetings
Thomas

Date: 16 Sep, 2004 on 10:21
Kim
Team member
in staff

View this member's profileSearch all posts from this memberSend an email to this memberhttp://www.haverblad.se
posts: 2128
since: 10 Dec, 2000
3. Re:IBM Releases Speech Technology To Open Source; Chance to Port to OS/2-eCS?
Reply to this topic with quote Modify your message
Tom,

I've used the speech tech within os2 many years ago, but mainly just to launch apps and utilities. But, I've also used the ViaVoice and it's an excellent software to use for alround to write short docs and launching apps, etc. As well used it to build voice macros, ie short voice command could launch multiply actions, etc.

Date: 16 Sep, 2004 on 10:53
Fahrvenugen
Premium member
in user

View this member's profileSearch all posts from this memberSend an email to this member
posts: 277
since: 10 Aug, 2004
4. Re:IBM Releases Speech Technology To Open Source; Chance to Port to OS/2-eCS?
Reply to this topic with quote Modify your message
I can't speak for others, but I have used VoiceType on occasion. True, it takes a bit to get used to, but can be quite useful I've found.

As for ViaVoice, while I welcome IBM open sourcing it, I also question whether it would be worthwhile to port to OS/2. I'd certainly encourage any programmers who wanted to take on the task, but on the other hand I would wonder if those programming skills could be better spent on some of the worthwhile projects which are reported to have a lack of developers.

Part of my reasoning comes from recalling a number of years ago when ViaVoice initially came out - hitting the Windows world and IBM promoted it heavily to an end-user market (well, heavily promoted for IBM standards). Despite getting good reviews it never really seemed to catch on to become a "must have" app. Other competing products came out around the same time with similar promotion (Dragonsoftware's Naturally Speaking, and I think there was a third but can't recall what it was called). The promotion even reached a point where applications such as Corel Word Perfect Office had a version which shipped with Speech Recognition (Corel's was essentially bundled with Naturally Speaking). Today I see little emphasis or promotion of any product in the speech recognition market. I hate to compare the OS/2 and Linux market with windows, but in this case I'd have to wonder - whether we like it or hate it, over 90% of PC users are on windows. If this market has relegated speech recognition to what appears to be essentially now a "niche" market, how much of a demand is there for an update to the speech engine in OS/2, where the user base is significantly less?

Of course, I'm just stating my opinion here, I don't have sales statistics to back up my statements - if someone has this info then I'd be happy to hear otherwise. I'm just commenting on my interpretation of what I've observed over the past 7 or 8 years.

-Fahr12

Date: 21 Sep, 2004 on 22:37
warpcafe
Premium member
in user

View this member's profileSearch all posts from this memberSend an email to this member
posts: 350
since: 26 Nov, 2002
5. Re:IBM Releases Speech Technology To Open Source; Chance to Port to OS/2-eCS?
Reply to this topic with quote Modify your message
Yes,

you have perfectly described the situation. I also doubt that it would make big sense to "waste" manpower in that niche, but on the other hand, speech technology will be the market of the future. Not on the PC maybe but in automotive it seems. The problem is that speech technology was not the "hot" topic for users in the last years. Instead, it was broadband, USB, file sharing, DVD and wireless. This also applies for users of OS/2 and eCS.
Maybe one day more users will see the benefit of speech technology, but unless there are "easy ways" like well-documented APIs for developers to integrate speech tech in their products, I see no solution at hand.
As far as I have understood, IBM didn't open-source the whole stuff but a subset and I don't know if this is "enough" to have an impact on community development.
I would love to have continuous speech recognition for textprocessing at least, but there is two sides in that: Speech tech *and* a text processing program that is enabled to use it. And this is where problems start. I would be satisifed if there is a text box that recognizes my speech and then I can paste this to the word processor, but that doesn't make really sense.
I was trying to use speech recognition for myself but I have lost interest due to the "learning curve" of both me and the software. Nevertheless, it is still very interesting and after all, aren't we somehow all fans of StarTrek, where speech recognition is present for almost every kind of device?

Greetings
Thomas

Date: 22 Sep, 2004 on 09:43
Kim
Team member
in staff

View this member's profileSearch all posts from this memberSend an email to this memberhttp://www.haverblad.se
posts: 2128
since: 10 Dec, 2000
6. Re:IBM Releases Speech Technology To Open Source; Chance to Port to OS/2-eCS?
Reply to this topic with quote Modify your message
Using ViaVoice takes as already mentioned a while to get the system to acknowledge and accept once voice. When using ViaVoice Gold, within work, I spent around 2 weeks to get it to accept my voice "errors" and got it to that point that I could navigate Windows 2000 without any problem and as well write letters without bigger issues. Main problem was usually that you needed a system with some nice cpu and ram capacity, even if the VV specs was set rather low. Latest product that I noticed has gain lots of users would be Philips voice software, but that has mainly been due to the fact of larger support for non-english users. Back in 1999 and 2000 IBM had plands to offer a broaden support; but as far I can see nothing much has happened. Guess that would be why IBM is now "giving" up some of the code to hopefully get some boost.
Date: 22 Sep, 2004 on 11:19
RobertM
Normal member
in user

View this member's profileSearch all posts from this memberSend an email to this member
posts: 152
since: 16 Aug, 2004
7. Re:IBM Releases Speech Technology To Open Source; Chance to Port to OS/2-eCS?
Reply to this topic with quote Modify your message
VoiceType has a continuous speech subsystem with swappable context related dictionaries - and also trains easier than ViaVoice - the differences in the back end perhaps the reason. Also, on faster machines the pause for discrete speech is miniscule - barely more than that of normal speech once VioceType is properly set up. One issue with proper setup is definitely microphone (and it's sensitivity to background and computer generated noises), gain settings, sound card noise sensitivity and the amount of noise the hardware generates. I've had better success using VoiceType on some old IBM PC Series machines than on numerous clones - though those IBMs came in nickel plated cases and were ridiculously grounded with very (electrically) quiet power supplies.

The ICSS (Subsystem) should work with ViaVoice's back end as the calls should be the same - it would just mean (if VoiceType Continuous Speech engine couldnt already handle it) that the speech engine could use larger dictionaries for continuous speech. The limitation was probably based on the small amount of memory the MWave Voice Cards used to come with (4MB I think) and bus speeds (transferring a 20,000 word recog. dict. to the MWave even if it had more than 4MB RAM via a 386 or 486 bus on an ISA connector would probably have taken a bit). I think that (no longer necessary) holdover when the engine was ported to OS/2 code to run on an Intel just remained or was never adressed. Additionally, the MWave did an excellent job at voice recognition in 486's (while only running at 25MHz), but Intel chips needed a lot of extra "horsepowere" to achieve the same results (over base OS requirements). Thats obviously largely because the MWave is a DSP - and thus designed for such things as far as was explained to me. IBM even released toolkits to allow any app (including dictation) to use the ICSS - and I'm not aware of any reason why they wouldnt work with the VoiceType Engine that came with Warp 4 as opposed to only with the MWave based VoiceType for OS/2. I dont recall the dictionary size limits, though there is the ability to switch dictionaries on the fly or subsets thereof if memory serves - as well as allowing audio out - which was a feature barely used in the GUI integration with Warp 4 (only used for VioceType error reponses instead of allowing the GUI to also use the MM Subsytem as usual).

I never did a full comparison of the APIs, but if they arent the same, they are very similar - and the beauty of Warp/eCS is the rest of the interaction is the GUI integration - ie: SOM and the WPS - which I think the current work on that might suffice.

IBM, btw, offered numerous continuous speech products that were OS/2 or OS/2 Warp based (mostly medical and legal fields for dictation). No idea how they worked and most or all were MWave based hardware running the engine interacting with OS/2 or based off of other DSPs (for phone systems, etc). As well as speech recognition, there were speech response systems like ScreenReader/2 and offerings from Elan - all decent sounding to very realistic - all discontinued and buried - the technology is all already there and quite capable and quite old and quite lost.

- Rob

Date: 23 Sep, 2004 on 06:29
Kim
Team member
in staff

View this member's profileSearch all posts from this memberSend an email to this memberhttp://www.haverblad.se
posts: 2128
since: 10 Dec, 2000
8. Re:IBM Releases Speech Technology To Open Source; Chance to Port to OS/2-eCS?
Reply to this topic with quote Modify your message
Two other interesting products within the IBM Speech series are (asume that it's still available) Via Voice Outloud that it's a speech software that reads out the text and this was used together with IBM HomePageReader (HPR) that "read" the content within homepages and used Outloud to read the content for blind users. The version I got only works with IE and I'm not sure if IBM has done any futher work on it since this section of IBM never gets out to speak about this kind of products.
Date: 23 Sep, 2004 on 07:05
IBM Releases Speech Technology To Open Source; Chance to Port to OS/2-eCS?
Post a new topic Reply to this Topic Printable Version of this Topic Forward this Topic to your Friend Topic Commands (for administrator or moderators only)
All times are CET+1. < Prev. | P. 1 | Next >
Go to:
 

Powered by UltraBoard 2000 Standard Edition,
Copyright © UltraScripts.com, Inc. 1999-2000.
Home | Gallery | Forums | Services | Webmail | Archive | Links | Contact Us | About Us
OS2 World.Com 2000-2004