Operations Summary: | |
float | getCompleteTimeout() |
float | getConfidenceLevel() |
float | getIncompleteTimeout() |
long | getNumResultAlternatives() |
float | getSensitivity() |
float | getSpeedVsAccuracy() |
boolean | isResultAudioProvided() |
boolean | isTrainingProvided() |
void | setCompleteTimeout(in float timeout) raises |
void | setConfidenceLevel(in float confidenceLevel) raises |
void | setIncompleteTimeout(in float timeout) raises |
void | setNumResultAlternatives(in long num) raises |
void | setResultAudioProvided(in boolean audio) raises |
void | setSensitivity(in float sensitivity) raises |
void | setSpeedVsAccuracy(in float speedVsAccuracy) raises |
void | setTrainingProvided(in boolean trainingProvided) raises |
Operations Inherited from EngineProperties | |
addPropertyChangeListener, getControlComponent, removePropertyChangeListener, reset |
Operations Inherited from Bonobo::Unknown | |
ref, unref, queryInterface |
Operation Details: |
Get the "CompleteTimeout" property.
setCompleteTimeout
Get the recognizer's "ConfidenceLevel" property.
setConfidenceLevel
Get the "IncompleteTimeout" property.
setIncompleteTimeout
Get the "NumResultAlternatives" property.
setNumResultAlternatives
Get the "Sensitivity" property.
setSensitivity
Get the "SpeedVsAccuracy" property.
setSpeedVsAccuracy
Get the "ResultAudioProvided" property.
setResultAudioProvided
Get the "TrainingProvided" property.
setTrainingProvided
Set the CompleteTimeout
property in seconds. This
timeout value determines the length of silence required following
user speech before the recognizer finalizes a result (with an
RESULT_ACCEPTED
or RESULT_REJECTED
event). The complete timeout is used when the speech is a
complete match an active RuleGrammar
. By contrast,
the incomplete timeout is used when the speech is an incomplete
match to an active grammar.
A long complete timeout value delays the result completion and therefore makes the computer's response slow. A short complete timeout may lead to an utterance being broken up inappropriately. Reasonable complete timeout values are typically in the range of 0.3 seconds to 1.0 seconds.
CompleteTimeout
property in seconds. getCompleteTimeout
Set the recognizer's "ConfidenceLevel" property. The confidence level value can very between 0.0 and 1.0. A value of 0.5 is the default for the recognizer. A value of 1.0 requires the recognizer to have maximum confidence in its results or otherwise reject them. A value of 0.0 requires only low confidence so fewer results are rejected.
getConfidenceLevel
Set the IncompleteTimeout
property in seconds.
The timeout value determines the required length of silence
following user speech after which a recognizer finalizes a result.
The incomplete timeout applies when the speech prior to the
silence is an incomplete match of the active
RuleGrammars
. In this case, once the timeout is
triggered, the partial result is rejected (with a
RESULT_REJECTED
event).
The incomplete timeout also applies when the speech prior to the silence is a complete match of an active grammar, but where it is possible to speak further and still match the grammar. For example, in a grammar for digit sequences for telephone numbers it might be legal to speak either 7 or 10 digits. If the user pauses briefly after speaking 7 digits then the incomplete timeout applies because the user may then continue with a further 3 digits.
By contrast, the complete timeout is used when the speech is a
complete match to an active RuleGrammar
and no
further words can be spoken.
A long complete timeout value delays the result completion and therefore makes the computer's response slow. A short incomplete timeout may lead to an utterance being broken up inappropriately.
The incomplete timeout is usually longer than the complete timeout to allow users to pause mid-utterance (for example, to breathe).
IncompleteTimeout
property in seconds. getIncompleteTimeout
Set the "NumResultAlternatives" property. This property indicates
the preferred maximum number of N-best alternatives in
FinalDictationResult
and FinalRuleResult
objects. A value of 0 or 1 indicates that the application does
not want alternatives; that is, it only wants the best guess.
Recognizers are not required to provide this number of
alternatives for every result and the number of alternatives
may vary from result to result. Recognizers should only provide
alternative tokens which are considered reasonable guesses: that
is, the alternatives should be above the
ConfidenceLevel
set through this interface (unless
the Result
is rejected).
Providing alternatives requires additional computing power. Applications should only request the number of alternatives that they are likely to use.
getNumResultAlternatives
FinalDictationResult#getAlternativeTokens
FinalRuleResult#getAlternativeTokens
Set the "ResultAudioProvided" property. If set to true, the
recognizer is requested to provide audio with
FinalResult
objects. If available, the audio
is provided through the getAudio
methods
of the FinalResult
interface.
Some recognizers that can provide audio for a
FinalResult
cannot provide audio for all results.
Applications need test audio available individually for results.
(For example, a recognizer might only provide audio for dictation
results.)
A Recognizer
that does not provide audio for any
results throws a IllegalArgumentException when an app attempts
to set the value to true.
isResultAudioProvided
FinalResult#getAudio
FinalResult#getAudioForToken
Set the "Sensitivity" property. The sensitivity level can vary between 0.0 and 1.0. A value of 0.5 is the default for the recognizer. A value of 1.0 gives maximum sensitivity and makes the recognizer sensitive to quiet input but more sensitive to noise. A value of 0.0 gives minimum sensitivity requiring the user to speak loudly and making the recognizer less sensitive to background noise.
Note: some recognizers set the gain automatically during use, or through a setup "Wizard". On these recognizers the sensitivity adjustment should be used only in extreme cases where the automatic settings are not adequate.
getSensitivity
Set the "SpeedVsAccuracy" property. The default value is 0.5 and is the factory-defined compromise of speed and accuracy. A value of 0.0 minimizes response time. A value of 1.0 maximizes recognition accuracy.
Why are speed and accuracy a trade-off? A recognizer determines what a user says by testing different possible sequence of words (with legal sequences being defined by the active grammars). If the recognizer tests more sequences it is more likely to find the correct sequence, but there is additional processing so it is slower. Increasing grammar complexity and decreasing the computer power both make this trade-off more important. Conversely, a simpler grammar or more powerful computer make the trade-off less important.
getSpeedVsAccuracy
Set the TrainingProvided
property. If true, request
a recognizer to provide training information for
FinalResult
objects through the
tokenCorrection
method.
Not all recognizers support training. Also, recognizers that do
support training are not required to support training data for all
results. For example, a recognizer might only produce training data
with dictation results. A Recognizer
that does not
support training raises a IllegalArgumentException when an
application attempts to set the value to true.
TrainingProvided
property. isTrainingProvided
FinalResult#tokenCorrection