Approximate Matching of Hand-Drawn Pictograms

We describe a new naming paradigm for pen-based computing which we call Pictographic Naming. Using our approach, traditional character-by-character handwriting recognition (HWX) is avoided. Instead, we use a combination of user interface conventions and approximate matching techniques. Since pictographic names incorporate pen-stroke data, they can never be reproduced exactly, so name lookup becomes a small-dictionary recognition problem. We give efficient algorithms for this problem, and show that we can accurately look up handwritten names in English and Japanese over a reasonably sized dictionary.