Good morning, My apologies if this isn't quite the right list for this question. We're in the middle of developing a VoIP IVR application and in one of our prompts we need the user to speak names of people. Once the recording is complete, we'd like to automatically convert that recording to text for saving to a database. Currently, we've been testing Google Voice's API and have had pretty decent luck (compared to other 'free' solutions); however, we've also encountered gems such as: the lord's johnson (Delores Johnson) oh my god it's got a carburetor (Ernesto Mogato, Edgar Carbrera) we bake a 10 year reunion (Willie Baker, Daneil Munjez) It appears that many of of the speech-to-text engines are designed for phrases instead of names. Would anyone happen to have any recommendations for speech-to-text engines that do a good job recognizing names? Are any of them better than others regarding dealing with accents? We're open to suggestions of the free, open-source, or commercial variety if the accuracy is there. Thank you, Brandon Lehmann
On 10/2/13 8:06 AM, Brandon Lehmann wrote:
It appears that many of of the speech-to-text engines are designed for phrases instead of names. Would anyone happen to have any recommendations for speech-to-text engines that do a good job recognizing names? Are any of them better than others regarding dealing with accents?
None will be perfect, and spelling of names varies. You'll probably have better luck going the other direction, typing the names into a text-to-speech application and having it speak the names. "The lord's johnson" - Heh! Of course, one can play games with text-to-speech conversion as well. I have a friend who reads email this way and will occasionally throw in a phrase like "If you see Kaye, why owe you?" just to keep him on his toes. -- Jay Hennigan - CCIE #7880 - Network Engineering - jay at impulse.net Impulse Internet Service - http://www.impulse.net/ Your local telephone and internet company - 805 884-6323 - WB6RDV
I'm sorry to hear that you're having trouble, but I think that I have my first entry in the "Best of VoiceOps Hall of Fame"! David From: VoiceOps [mailto:voiceops-bounces at voiceops.org] On Behalf Of Brandon Lehmann Sent: Wednesday, October 02, 2013 08:06 To: voiceops at voiceops.org Subject: [VoiceOps] (no subject) Good morning, My apologies if this isn't quite the right list for this question. We're in the middle of developing a VoIP IVR application and in one of our prompts we need the user to speak names of people. Once the recording is complete, we'd like to automatically convert that recording to text for saving to a database. Currently, we've been testing Google Voice's API and have had pretty decent luck (compared to other 'free' solutions); however, we've also encountered gems such as: the lord's johnson (Delores Johnson) oh my god it's got a carburetor (Ernesto Mogato, Edgar Carbrera) we bake a 10 year reunion (Willie Baker, Daneil Munjez) It appears that many of of the speech-to-text engines are designed for phrases instead of names. Would anyone happen to have any recommendations for speech-to-text engines that do a good job recognizing names? Are any of them better than others regarding dealing with accents? We're open to suggestions of the free, open-source, or commercial variety if the accuracy is there. Thank you, Brandon Lehmann This message and any attachments are intended only for the use of the addressee and may contain information that is privileged and confidential. If the reader of the message is not the intended recipient or an authorized representative of the intended recipient, you are hereby notified that any dissemination of this communication is strictly prohibited. If you have received this communication in error, please notify us immediately by e-mail and delete the message and any attachments from your system.
You are very welcome :D We?ve had a lot of fun reading what the system is picking up. From: Hiers, David [mailto:David.Hiers at adp.com] Sent: Wednesday, October 02, 2013 3:16 PM To: Brandon Lehmann; voiceops at voiceops.org Subject: RE: [VoiceOps] (no subject) I'm sorry to hear that you're having trouble, but I think that I have my first entry in the "Best of VoiceOps Hall of Fame"! David From: VoiceOps [mailto:voiceops-bounces at voiceops.org] On Behalf Of Brandon Lehmann Sent: Wednesday, October 02, 2013 08:06 To: voiceops at voiceops.org Subject: [VoiceOps] (no subject) Good morning, My apologies if this isn't quite the right list for this question. We're in the middle of developing a VoIP IVR application and in one of our prompts we need the user to speak names of people. Once the recording is complete, we'd like to automatically convert that recording to text for saving to a database. Currently, we've been testing Google Voice's API and have had pretty decent luck (compared to other 'free' solutions); however, we've also encountered gems such as: the lord's johnson (Delores Johnson) oh my god it's got a carburetor (Ernesto Mogato, Edgar Carbrera) we bake a 10 year reunion (Willie Baker, Daneil Munjez) It appears that many of of the speech-to-text engines are designed for phrases instead of names. Would anyone happen to have any recommendations for speech-to-text engines that do a good job recognizing names? Are any of them better than others regarding dealing with accents? We're open to suggestions of the free, open-source, or commercial variety if the accuracy is there. Thank you, Brandon Lehmann _____ This message and any attachments are intended only for the use of the addressee and may contain information that is privileged and confidential. If the reader of the message is not the intended recipient or an authorized representative of the intended recipient, you are hereby notified that any dissemination of this communication is strictly prohibited. If you have received this communication in error, please notify us immediately by e-mail and delete the message and any attachments from your system.
----- Original Message -----
From: "Brandon Lehmann" <brandon at bitradius.com>
oh my god it's got a carburetor (Ernesto Mogato, Edgar Carbrera)
That just made my whole day worthwhile. Thank you. :-) Now that we've gotten the humor part of the day out of the way, I'm pretty sure that what you want to do is -- while not impossible -- something you can only ever expect to get 50-70% reliability out of, if that high. Proper names are only barely required to follow the usual pronunciation rules of any given language, and are prone to come from whatever language was native to the person's parents, which could be anything (there are somewhere over 4000 langauges on the planet, IIRC). Cheers, -- jra -- Jay R. Ashworth Baylink jra at baylink.com Designer The Things I Think RFC 2100 Ashworth & Associates http://baylink.pitas.com 2000 Land Rover DII St Petersburg FL USA #natog +1 727 647 1274
On Oct 2, 2013, at 3:55 PM, Jay Ashworth wrote:
Now that we've gotten the humor part of the day out of the way, I'm pretty sure that what you want to do is -- while not impossible -- something you can only ever expect to get 50-70% reliability out of, if that high.
Proper names are only barely required to follow the usual pronunciation rules of any given language, and are prone to come from whatever language was native to the person's parents, which could be anything (there are somewhere over 4000 langauges on the planet, IIRC).
I've seen Nuance do some pretty amazing stuff in call center applications. http://www.nuance.com/for-business/by-solution/customer-service-solutions/so... But I doubt it's cheap, judging by the kind of cars my friend (who does the integration work) drives. --Chris
The powers that be made the comment today that "Dragon can do it, why can't you?" I'm willing to be that's out of the budget but I'll definitely make a phone call.
-----Original Message----- From: VoiceOps [mailto:voiceops-bounces at voiceops.org] On Behalf Of Chris Boyd Sent: Wednesday, October 02, 2013 5:43 PM To: VoiceOps Subject: Re: [VoiceOps] (no subject)
On Oct 2, 2013, at 3:55 PM, Jay Ashworth wrote:
Now that we've gotten the humor part of the day out of the way, I'm pretty sure that what you want to do is -- while not impossible -- something you can only ever expect to get 50-70% reliability out of, if that high.
Proper names are only barely required to follow the usual pronunciation rules of any given language, and are prone to come from whatever language was native to the person's parents, which could be anything (there are somewhere over 4000 langauges on the planet, IIRC).
I've seen Nuance do some pretty amazing stuff in call center applications.
http://www.nuance.com/for-business/by-solution/customer-service- solutions/solutions-services/inbound-solutions/self-service- automation/recognizer/index.htm
But I doubt it's cheap, judging by the kind of cars my friend (who does the integration work) drives.
--Chris
_______________________________________________ VoiceOps mailing list VoiceOps at voiceops.org https://puck.nether.net/mailman/listinfo/voiceops
In all honesty, I informed the powers that be that we'd be lucky to hit 25% reliability; however, we all know how suits work. I'm glad I'm hearing the same thing I told them in the beginning. I was starting to think that maybe they were right and I was crazy.
-----Original Message----- From: VoiceOps [mailto:voiceops-bounces at voiceops.org] On Behalf Of Jay Ashworth Sent: Wednesday, October 02, 2013 4:55 PM To: voiceops at voiceops.org Subject: Re: [VoiceOps] (no subject)
----- Original Message -----
From: "Brandon Lehmann" <brandon at bitradius.com>
oh my god it's got a carburetor (Ernesto Mogato, Edgar Carbrera)
That just made my whole day worthwhile. Thank you. :-)
Now that we've gotten the humor part of the day out of the way, I'm pretty sure that what you want to do is -- while not impossible -- something you can only ever expect to get 50-70% reliability out of, if that high.
Proper names are only barely required to follow the usual pronunciation rules of any given language, and are prone to come from whatever language was native to the person's parents, which could be anything (there are somewhere over 4000 langauges on the planet, IIRC).
Cheers, -- jra -- Jay R. Ashworth Baylink jra at baylink.com Designer The Things I Think RFC 2100 Ashworth & Associates http://baylink.pitas.com 2000 Land Rover DII St Petersburg FL USA #natog +1 727 647 1274 _______________________________________________ VoiceOps mailing list VoiceOps at voiceops.org https://puck.nether.net/mailman/listinfo/voiceops
Ask them how software is expected to differentiate between Marques/Marcus, Stephen/Steven, 4 flavors of Britteney and my favorite Tiphphany/Tiffany. -Owen ----- Original Message ----- From: "Brandon Lehmann" <brandon at bitradius.com> To: "Jay Ashworth" <jra at baylink.com>, voiceops at voiceops.org Sent: Wednesday, October 2, 2013 7:27:51 PM Subject: Re: [VoiceOps] (no subject) In all honesty, I informed the powers that be that we'd be lucky to hit 25% reliability; however, we all know how suits work. I'm glad I'm hearing the same thing I told them in the beginning. I was starting to think that maybe they were right and I was crazy.
-----Original Message----- From: VoiceOps [mailto:voiceops-bounces at voiceops.org] On Behalf Of Jay Ashworth Sent: Wednesday, October 02, 2013 4:55 PM To: voiceops at voiceops.org Subject: Re: [VoiceOps] (no subject)
----- Original Message -----
From: "Brandon Lehmann" <brandon at bitradius.com>
oh my god it's got a carburetor (Ernesto Mogato, Edgar Carbrera)
That just made my whole day worthwhile. Thank you. :-)
Now that we've gotten the humor part of the day out of the way, I'm pretty sure that what you want to do is -- while not impossible -- something you can only ever expect to get 50-70% reliability out of, if that high.
Proper names are only barely required to follow the usual pronunciation rules of any given language, and are prone to come from whatever language was native to the person's parents, which could be anything (there are somewhere over 4000 langauges on the planet, IIRC).
Cheers, -- jra -- Jay R. Ashworth Baylink jra at baylink.com Designer The Things I Think RFC 2100 Ashworth & Associates http://baylink.pitas.com 2000 Land Rover DII St Petersburg FL USA #natog +1 727 647 1274 _______________________________________________ VoiceOps mailing list VoiceOps at voiceops.org https://puck.nether.net/mailman/listinfo/voiceops
_______________________________________________ VoiceOps mailing list VoiceOps at voiceops.org https://puck.nether.net/mailman/listinfo/voiceops
participants (6)
-
brandon＠bitradius.com -
cboyd＠gizmopartners.com -
David.Hiers＠adp.com -
jay＠west.net -
jra＠baylink.com -
owen＠impulse.net