C H A P T E R 9
Using the Recognition System
When troubleshooting recognition errors, remember that view flags may enable
multiple dictionaries and that the sets of dictionaries enabled by various flags
As a general rule, the fastest and most accurate recognition occurs when the fewest
recognizers and dictionaries necessary to successfully analyze the input are
enabled. Enabling unnecessary recognizers and dictionaries may decrease the
speed and accuracy with which recognition is performed.
flags enable text recognition in
views that accept pen input. Either flag enables the text recognizer specified by the
Handwriting Recognition preferences slip.
Each of these flags specifies different recognition options and dictionary sets. The
unique behaviors associated with each flag are demonstrated most clearly by the
cursive recognizer. The cursive recognizer can be made to return only words
present in the set of dictionaries available to the view performing recognition. In
contrast, the printed recognizer can always return words or letter combinations that
are not present in dictionaries.
flag enables a default set of dictionaries that provide
vocabulary used in common speech, names of days, names of months, proper
names, and words in the user dictionary. When the
flag is set
flag is not, the cursive recognizer returns only words
that appear in the set of dictionaries available to the view performing recognition.
Note that the complete set of dictionaries available to the view may include those
enabled by other flags. For example, the NTK view editor provides a Field Type
popup menu that allows you to specify whether the view is to accept phone, date,
time, address or name data. The choices in this menu set the
respectively. Each of these flags enables one or more dictionaries suited to
recognizing the specified input data. Custom dictionaries may also be made
available to the view performing recognition by setting the
flag and providing a valid
slot in the
view that performs recognition.
flag enables the cursive recognizer's letter-by-letter
recognition option. When the
flag is set, the cursive recognizer
may return words not appearing in dictionaries as well as nonword letter
combinations. Note that this configuration increases the cursive recognizer's chances
of misrecognizing words that appear in the set of dictionaries available to it.
Although both text recognizers provide a letter-based recognition feature, the two
recognition engines are completely distinct. Consequently, the results produced by