This is a fork of the Dragonfly project, a speech recognition framework for Python. The originating project source can be found at dictation-toolbox/dragonfly.

This fork is intended to accommodate specific modifications required for applications built by JWebmeister. Modifications include:

• Re-enabled notifications to KaldiRecObsManager within the Kaldi engine backend.
• Added notify_partial_recognition and related to engine and recobs, specifically for the Kaldi engine.
• Added listen_key (keyboard or mouse) with hold or toggle modes to Kaldi engine. Hold or toggle modes include:
  • 0: Hold; listen only while key is pressed, release key to finish utterance.
  • 1: Toggle; listen while on, press key to toggle on or off, must toggle off to finish utterance.
  • 2: Global toggle; always listening while on, press key to toggle on or off, finishes utterance based on Voice Activity Detector (VAD).
  • -1: Priority hold; always listen for only priority grammar (and all recobs), listen for all grammar while key is pressed, release key to finish utterance and also deactivate non-priority grammar.
• Modified on_recognition (and related) to include (words, results, rule, node)

These modifications are intended to:

• decrease the delay between speech and the execution of matching rules callback functions.
  • e.g. "Freeze this is the police!", match on recognition of the word "Freeze", and execute a function before the rest of the sentence has been fully spoken.
• reduce false positive recognitions.
  • e.g. change listening mode to on or off with a key press.

First, clone this repo into a folder, e.g. dragonfly/.

Run the following command (or similar) while in the dragonfly project folder:

pip install -e .

To build the dragonfly python package, run these commands in the projects root directory.

pip install build  
python -m build

For further information on how to use Dragonfly:

• source code for this fork at jwebmeister/dragonfly.
• documentation is available online at Read the Docs.
• the originating project source can be found at dictation-toolbox/dragonfly.