From what I gathered from the article, using a method like that in Symbolism (at least for the time being) wouldn't be possible. They seem to be using a time sequence based ("on-line") algorithm, where the order of strokes seem to matter. With Symbolism, we (have to?) use more traditional OCR ("off-line") methods for matching images, that can match images as a whole, with no idea about order of strokes, pen-lifting, etc.