C H A P T E R 9
About the Recognition System
Recognition may fail when the handwritten input is too sloppy for the system to
make a good match against its internal handwriting model, when the view is not con-
figured correctly for the intended input, or (in the case of dictionary-based recognition
only) when none of the interpretations of the input strokes match a dictionary entry.
In such cases, the recognition system may return sketch ink or ink text.
Ink text looks similar to sketch ink; however, ink text is scaled and placed in a
view as text. Sketch ink is not placed in a paragraph but
drawn in a
view on top of anything else that appears in the
parent. Both ink text and sketch ink hold stroke data
that can be used to recognize the strokes at another time. Deferred recognition--
the process of recognizing saved ink at a later time--is described in more detail in
"Deferred Recognition" (page 10-5), in Chapter 10, "Recognition: Advanced
The system supplies a variety of dictionaries against which names, places, dates,
times, phone numbers, and commonly used words are matched. There are two
kinds of dictionaries used for text recognition: enumerated and lexical. An
enumerated dictionary is simply a list of strings that can be matched. A lexical
dictionary specifies a grammar or syntax that is used to classify user input. The
kind of dictionary used for a particular task is dependent upon task-specific
requirements. For example, it would be impractical to create an enumerated
dictionary of phone numbers; however, the clearly defined format imposed on these
numbers makes them ideal candidates for definition in a lexical dictionary.
The specific set of dictionaries that the system provides for a particular purpose
generally varies according to the user's locale. For example, because currency
formats vary from country to country, the particular lexical dictionary that the
system uses for matching monetary values may change according to the current
locale. However, you usually need not be concerned with the specific set of
dictionaries used by a particular locale. For more information, see Chapter 20,
"Localizing Newton Applications."
Dictionaries can be in ROM or in RAM (internal or card-based). Most of the system-
supplied dictionaries are in ROM; however, the user dictionary resides in RAM.
Applications must never add items to the user dictionary without the user's
consent. The user dictionary is intended to be solely in the user's control--adding
items to it is akin to changing the user's handwriting preferences or Names entries.
It's also important to leave room for users to store their own items.