Now here's a piece of joined-up thinking:
You know that anti-spam thing - where you have to type in a random set of letters and numbers in an obscured image? Oh, go on .. you do. Well, some clear-thinking bods at Carnegie Mellon Uni in Pittsburgh have found a use for it. Translation from printed word to computer data.
The "random" text used in this idea is in fact the image of a word which cannot be confidently interpreted by OCR. This is already distorted - by age and distortion - and random "noise" is added for good measure. A control word is added, giving two words to "translate". The user then types in both words; if the control word is entered correctly, the translated word is stored.
By randomly assigning words to the millions of users online every day, the translation process is delegated to the world in general. Bit by bit, word by word, old texts are being computerised for posterity. All in the name of security. And, for free.
Everyone's a winner!
No comments:
Post a Comment