Rubber Bucket

Optical character recognition for mobile phone cameras. Why? Because, used in conjunction with basic machine translation or dictionary lookup,  you have real life, with subtitles.

Imagine this: you’re in a restaurant in a foreign country. They don’t have an English menu, and the waiter’s English isn’t great either. Pull out your mobile, switch on the camera viewfinder, hold it over each menu item in turn. As the phrase queso de cabra appears in the centre of the picture, an eggtimer appears in the corner of the screen, indicating that it’s attempting to recognise the words. A moment later, at the bottom of the screen in a shaded area appears the text queso de cabra in blue, and next to it, goat’s cheese in yellow. (The blue text confirms the original, so you can be sure it’s looking at the same words as you. The yellow text stands out more clearly as the info you really need.)

Optionally, it might also pronounce the original word and/or translation. Useful for learning the language.

Also handy for deciphering road signs, advertisements, notices in shops…

Advantages over electronic phrasebooks:

• It’s software - no expensive qwerty keyboard etc - therefore huge price advantage over phrasebooks.
• Enables easy downloads  - of updates, but also of additional language packs.
• Could detect diacritical marks or potentially even other scripts (e.g. Arabic, Japanese), whereas many English speakers would not know how to input non-English characters on a keyboard.
• Faster to use, more discrete, one less gadget to carry around or lose.

Potential issues:

• A reasonably high resolution camera would make it more useful, by allowing the user to zoom in on distant signs while maintaining a good enough resolution to allow OCR.
• A camera that could produce a reasonably sharp image down to distances of around 5cm, would be more useful. This would enable it to work with typefaces of preferably down to 12pt but at least 14pt. Smaller typefaces are probably unrealistic. This probably requires autofocus rather than the fixed focus lens found on most current mobiles.
• It would be very elegant, from the user’s point of view, not to require pre-selecting a special mode to use this function. In other words, any time the camera is on, it is attempting to detect and recognise text. However this is probably not possible with 3rd party software; it might only be an option for the phone manufacturer.
• It might be difficult to enable it to work on a live picture (possibly require image-stabilisation software?). Early/cheap versions might require the user to take a snapshot of the text before processing it.