The Speech Accent Archive at George Mason University is a neat idea. But it seems to be based on the premise of providing one exemplar of each place of origin -- for countries where there are multiple speakers, each one is identified as being from a unique city or town. This makes it less than optimally useful for studies of variable phenomena, for studies of things that depend on level of experience with English, and so on.
It'd also be nice to be able to get the audio in a convenient form, for further analysis. The quicktime .mov format in which they're stored is not among the more widely recognized formats, at least by audio analysis programs.
Posted by Mark Liberman at March 28, 2004 04:54 PM