Seems strange that anyone would prefer to write laboriously by hand for an idiot machine that can only understand characters when each one is formed clearly and regularly, when they could use a keyboard, but there is a widespread belief that many people do, and the US National Institute of Standards & Technology reports that mathematician Charles Wilson has developed new letter recognition software that is quick and accurate, enabling computer recognition in 4mS with 80% accuracy on unconstrained hand printed characters and greater than 99% accuracy on machine print; Wilson and his team worked with the US Census Bureau to collect 2,000 handwriting samples which are now part of a database that contains 294,000 numbers and 728,700 letters, which will be used as a standard to test new recognition software and hardware under development; the algorithms for Wilson’s recognition system are based on neural network technology and National Institute researchers are now working with the Internal Revenue Service to develop a model handprint recognition system – aw, c’mon – the clarity of one’s handwriting is in direct proportion to the pleasurability of the task, and isn’t everybody’s tax return right at the borderline of legibility?
