U. F.
The recognition system in the American Newton uses a multiple recognizer architecture, with distinct recognizers for words, shapes (graphics), and editing and command gestures. The word recognizer uses built-in dictionaries to choose words that, with sufficient probability, match what you write.
The recognition process involves a hierarchy of recognizers that take specific types of recognition units and produce more complex units as output. Each recognizer processes its input according to a set of rules, adds interpretations to the input, and combines this more refined information in an output recognition unit that is then passed on to higher-level recognizers. To recognize a word that is made up of printed, separate characters, the strokes that make up the word are interpreted as cha
racter parts, then the parts as characters, and then the characters as a word.
This structure lets developers add new recognizers at different levels of the hierarchy. This part of the Newton operating system is only accessible through the C/C++ interface. At this point, the NewtonScript programming language and the NTK (Newton Tool Kit) don't offer tools to add recognizers to the system.
Because Americans and Germans are taught different handwriting styles at an early age, certain elements of the Newton recognition code were modified for the German version. These modifications include a combined printed and cursive handwriting recognizer, instead of two separate recognizers as for the American version. Apple declined to disclose any other modified elements.
Illustration: Strokes
Newton Recognition Architecture
