Human identification of letters in mixed-script handwriting: anupper bound on recognition rates

摘要

This paper focuses on a reading task consisting of thenidentification of letters in mixed-script handwritten words. This tasknis performed by humans using extended or limited linguistic context.nTheir performance rate is to give an upper bound on recognition rates ofncomputer programs designed to recognize handwritten letters innmixed-script writing. Many recognition algorithms are being developed innthe research community, and there is a need for establishing ways toncompare them. As some effort is on its way to give large test sets withnstandard formats, we propose an algorithm to determine a test set ofnreduced size that is appropriate for the task to achieve (the type ofntexts or words to be recognized). Also, with respect to a particularntask, we propose a method for finding an upper limit to the letternrecognition rate to aim for