Point, Shoot and Translate into English
edstromp points out this New York Times "story on using a pocket pc to translate a street sign. It requires at least a dialup connection as it sends the photo to a server for the majority of the processing: OCR, translation, English overlay for new image, and then transmission back to the user. All said and done, it takes about 15 seconds to translate a street sign. Put this with some augumented reality, and you have a rather useful tool."
OK, maybe translating train schedules and restaurant menus is good. But street signs, especially, are supposed to be unambiguous, their meaning readily apparent to anyone, whether literate in their native language or not.
And does this thing work on signs that some redneck has shot holes in with a 12-gauge?
Why not allow the OCR program, and any necessary foreign language translation dictionaries to live in the PDA's memory? I can't see it taking up anymore than several MB, (which could certaintly be offloaded when not traveling).
Is there something I really don't understand here?