after each hypothesis, we try to find one (or more) "commands",
which we define as "(contentlessWord | modifierWord)* actionWord",
i.e. a sequence of words leading up to and including an actionWord.
Some Sphinx frontend processors that come in handy when working
with Sphinx and InproTK:
An improved implementation of Sphinx' Speech Marker with much
improved incremental behaviour (QuickSpeechmarker.java)
A processor that reads audio from an RTP stream
(RtpRecvProcessor.java)
A processor that splits off all frontend processing into a
separate thread (ThreadingFrontendBuffer.java)
A processor that logs passing audio to TEDview
A simple wrapper for both system and user uttered installments
- for user installments references to recognized WordIUs are kept
used to infer the spoken text in toPayload()
- for system installments a reference to the corresponding dialogueAct
is kept and the spoken text is stored in the variable @see{tts}.
call this, if you want to provide a sameLevelLink and a groundedIn list
and you want groundedIn to be deeply SLLed to the sameLevelLink's groundedIn-IUs
An IUDocument stores a list of current IUs,
edits since the last update and
the string for the (partial) next IU
It handles HypothesisChangeListeners
which are notified, when the IUList changes (or is committed)
(and can be set via setListeners())
IUs are committed and the list is reset after an explicit call to commit()
(when used as the document of a JTextField, this can be done by
calling commit() from a JTextField's ActionListener)
A simple constructor initiating an empty information state and
a generic top-down search update mechanism producing GROUND,
CLARIFY and REQUEST output.