Has the industry abandoned prompt concatenation? I mean, yes, you can hear recorded voices stumbling over numbers and words on just about any system you tune into, but is anyone trying to make this better still, or is it just me? Is the industry just waiting for tts to become tolerable enough to use fulltime, or for users to get used to broken speech enough to accept it? This is how I describe my job at cocktail parties when asked: “You know those systems you call and talk to a computer instead of a person, like the phone company?“ <insert long pause for customary rolling of the eyes or grunt of disgust> “Well, I make them better.” <smile>. I’ve even read research lately where some would say it’s easier on the ears to hear a prompt entirely in tts rather than with as much recorded as possible and only the most dynamic data in tts. What burns me is that I’ve worked with the best in the business in this area and saw her talent wasted by bureaucracy and lack of vision. If we want it, it’s attainable – high quality, pre-recorded, concatenated speech for much more content than we are, well, content with now. Please let me help you help you.
Simonie
July 10th, 2007
3:25 pm
Me. Talk. Pretty. One. Day.
This issue drives me nuts too. And the thing is, if you do a good job with the recordings and programming, it doesn’t have to be like this - it can sound really good - much better than the current TTS. It’s such a balancing act, though, especially with variable data like names of streets, people, or places - you can’t always record them all. So, not only do you have to worry about the prosody of the recordings, you have to worry about integrating the TTS with the recordings so they sound fairly natural.
-Sarah Wayland
July 19th, 2007
6:58 am
Agreed! There is a lot to balance in VUI design. I’m just surprised that after more than 10 years in this business, people still accept poor quality output as long as the underlying app can dazzle. I’m very user-centric and know that as an industry, we’re still the butt of a lot of jokes because of this exact issue, that has already been solved!