Abstract
Overlap errors, in which two keys are pressed down at once, are a common typing error for people with motor disabilities. Keyguards are a commonly suggested means to reduce overlap errors. However, they are also unpopular with many users. We present an alternative to the keyguard, a software filter which targets overlap errors. Basic, keystroke timing-based, and language-based techniques for identifying and correcting overlap errors are described. Their performance is compared using a corpus of typing data recorded by keyboard users with motor disabilities. The best filter performance was obtained by using keystroke timing characteristics to identify and filter out extra characters. Accuracy of error identification was dependent on the typing style of the user. The filter accurately corrected 80% of the overlap errors presented. Combining the identification and correction techniques gave a 50-75% reduction in errors for the three study participants with the highest error rates.