I think voice recognition, especially when combined with dynamic, data-driven systems is a Big Thing. That said, we have a long way to go as demonstrated by this slightly edited Google Voice transcript of its interaction with an automated survey app from a car dealer:
We are calling on behalf of <dealer> to. Sorry I didn’t hear that. Please listen for instructions following this message. We are calling on behalf of <dealer> to ask a few questions about the quality of service you received. There are 5 questions and it should only take a minute of your time. Please answer yes or no to the following question when arriving for service where you graded greeted promptly. Sorry I didn’t hear that. Please say yes or no, when arriving for service where you graded greeted promptly. Sorry I didn’t hear that. Please Press 1 for Yes, or two for no. Sorry i’m having such a difficult time understanding you. Would you like to continue? Please Press 1 for Yes, or two for no. Sorry I didn’t hear that. Would you like to continue? Please Press 1 for Yes, or two for no. Sorry i’m having such a difficult time understanding you. You can contact us AT (508) ***-**** We appreciate your time and thank you for choosing <dealer> goodbye.
I wish more startups were doing cool things with voice and speech rec, especially now that services like Twilio are available..
Voice services: voice recognition depends heavily on the language it expects to get as input. and on the language the system should use for the output. As long as language preferences are not transmitted during the call set-up process, automatic voice systems are limited in their use.
Good point, Christoph. Do you know of a protocol for this or of a company that has implemented this?
There is nothing like this on the market and no standardized protocols. It is actually quite challenging to do as it has to work in different scenarios (ITU-based and non-ITU-based), so that it also works across carriers and standardized and proprietary VoIP systems. It would be definitively worth doing it, though.