Estimating word length for efficient text analysis

Although the problem of recognising machine printed words has been largely solved, no ideal solution has been found for the problem of handwritten word recognition, and especially cursive handwritten word recognition. This paper describes a simple technique for estimating word length in unconstrained text which offers fast implementation without a high memory and computational overhead. Such a technique has potential applications in a range of tasks, and has been evaluated specifically in the context of improving the performance in reading the legal amount field on British bank cheques.