My very unscientific opinion is that Google does fairly well with different accents. Always a matter of degree of course. Speaking in complete grammatical sentences
is more important regardless (which is why they don’t always do that well with conversational transcription).
The button is helpful, but a tablet times out if it does not hear anything for a bit. Google docs just did the same, and a mouse click was required to get it
going again..
We just turn the actual mic on/off with a button on it's cable when we want to stop transcription.
Switching between accents of different speakers I think will be the challenge. I don't think Google Speech has any training mechanism. I was hoping we could train a system to recognize the substitutions
it needs to do in order to correctly transcribe Tagalog accented English and then it will work with both American and Tagalog accented English. I hope we can teach one of these programs the correct substitutions for whole common words but I would not want the
program to generalize the rules and convert all instances of a phoneme production. That would likely wreck its ability to transcribe American accented English.
I have asked his wife to download Google Speech Recognition on her Surface and give it a spin.
I welcome all other thoughts.
Sincerely,
Amy Roman, MS, CCC-SLP
Augmentative Communication Specialist
Forbes Norris ALS Research and Treatment Center
2324 Sacramento Street
San Francisco, CA 94115
Cell (415)518-0592
Fax (415)600-3778
I tried the dragon dictation app on my iPhone and it worked well with different voices, like Ed’s comment the user would need to hit record and done.
Richard Hurtig, Ph.D.
ASHA Fellow
Professor Emeritus
Department of Communication Sciences & Disorders
The University of Iowa
Google speech will work with many different voices, vs WSR(Windows Speech Rec) or DNS (Dragon NaturallySPeaking) that are trained
to one particular voice. DNS actually may work this way, but is not designed to do so…
I did not think about trying it in google docs, on the computer. On my phone, I experimented a bit; my wife and I could both “talk”
and it transcribed what we said.
One additional issue across the board. The button to start SR needs to be pushed in both cases (touch screen on tablet, or mouse
click on Computer). SO a caregiver may need to do that….
Thanks so very much for your recommendations.
Since they already have a Windows 10 tablet, would you recommend we try Google speech recognition on it or do you think it works better (or only works) on an Android?
So just to be sure I am understanding.... the advantages of Google speech recognition is it picks up many different voices better than some of the other programs and the
advantage to Android is the Bluetooth mic option?
Sincerely,
Amy Roman, MS, CCC-SLP
Augmentative Communication Specialist
Forbes Norris ALS Research and Treatment Center
2324 Sacramento Street
San Francisco, CA 94115
Cell (415)518-0592
Fax (415)600-3778
So if I was only going to use speech recognition program on the PC, I would probably go with Dragon over Windows speech recognition at this point. Hands down.
As noted, I have seen Windows speech recognition to be kind of buggy. And you can buy Dragon 13 home version which is still more functional than Windows speech recognition for around 35 or $40 on Amazon. All of that being said, any speech recognition for the
computer is going to require fairly specific dictation skills that do not usually lend themselves to conversations, especially in a group. Also, Windows speech recognition and Dragon are both going to be trained to an individual speaker, you would not be able
to pass them around.
If I were you guys I would go with a android tablet and use Google speech recognition into a memo pad. You can route audio from a Bluetooth microphone. Additionally,
Google speech recognition will pick up conversations and comments from multiple speakers with different voices. I doubt that it will do it all at once, people need to be careful to speak one at a time into the microphone. Also, it will help a lot if they do
use punctuation and try to stick with relatively clear speech.
To my knowledge, Siri is going to be speaker trainable, thus not as good for this purpose. Additionally given some recent research I did it sounds like you
cannot route audio through a Bluetooth headset for speech recognition purposes on iOS.
As a speech therapist I don't do as much work with speech recognition as many of you. I could use some guidance for some challenges I am having with one patient.
I am working with a man who had brainstem stroke. He is cognitively and linguistically intact but became deaf, paralyzed and non-verbal when he had a stroke. His eye movement
is severely reduced too and one eye was sewn shut since he was not blinking adequately to prevent ulceration.
For speech generation and computer access on a Zuvo SGD, I set him up with auto scanning which he can do with the slightest movement of his thumb.
In order for others to communicate with him we were using writing on a BoogieBoard but this is laborious and slow. I clamped the family's Surface with Windows 10 to the
mount for his Zuvo (using the JoyFactory clamping tablet mount which is affordable and good) and turned on Windows Speech Recognition and Notepad. We are using a hands free headset mic with a USB connection to the Surface for dictation.
His wife and visitors speak with Tagalog accented English. His kids speak with American accented English. The system initially did not recognize much of his wife's speech
but after I had the her go through the brief voice training for the Windows Speech Recognition software it worked much better. It was faster and easier than writing for her. She is the only one using the system at this point.
We are having a few challenges. Here are our questions.
1. The wife says every time she uses the Windows Speech Recognition it acts as if she has not done the training and she has to do it again. Is this normal for this program?
2. She is willing to buy Dragon if it will work better. All we are doing is text entry. Is it worth the upgrade?
3. Is there a system that works best when multiple speakers are using the Speech Recognition within the same session?
4. Would a wireless mic that could be passed around to various conversation partners in the room work as well as a connected USB mic?
5. Do people think the technology is good enough to have several speakers all understood? (with the same accent? with differing accents?)
Sincerely,
Amy Roman, MS, CCC-SLP
Augmentative Communication Specialist
Forbes Norris ALS Research and Treatment Center
2324 Sacramento Street
San Francisco, CA 94115
Cell (415)518-0592
Fax (415)600-3778
Confidentiality Notice: This message and any attachments are only for the intended recipient(s) and may contain confidential, privileged
and/or protected health information. If you are not the intended recipient of this message, please: 1) be advised that unauthorized review, use, copying, disclosure, or distribution is strictly prohibited and may be unlawful; 2) notify the sender of the delivery
by reply e-mail or contact Shirley Ryan AbilityLab's Privacy Officer at 312.238.0766 or xxxxxx@sralab.org;
and 3) delete and destroy all copies of the message and its attachments. Thank you.
Confidentiality Notice: This message and any attachments are only for the intended recipient(s) and may contain confidential, privileged and/or protected health information.
If you are not the intended recipient of this message, please: 1) be advised that unauthorized review, use, copying, disclosure, or distribution is strictly prohibited and may be unlawful; 2) notify the sender of the delivery by reply e-mail or contact Shirley
Ryan AbilityLab's Privacy Officer at 312.238.0766 orxxxxxx@sralab.org;
and 3) delete and destroy all copies of the message and its attachments. Thank you.
Confidentiality Notice: This message and any attachments
are only for the intended recipient(s) and may contain confidential, privileged and/or protected health information. If you are not the intended recipient of this message, please: 1) be advised that unauthorized review, use, copying, disclosure, or distribution
is strictly prohibited and may be unlawful; 2) notify the sender of the delivery by reply e-mail or contact Shirley Ryan AbilityLab's Privacy Officer at 312.238.0766 orxxxxxx@sralab.org;
and 3) delete and destroy all copies of the message and its attachments. Thank you.