I'm pleased to announce that the methods described in our recent NAACL paper [1] to filter intermediate ASR results and generate incremental ASR output are now implemented as a Sphinx-4 ResultListener. The code is available at http://www.ling.uni-potsdam.de/~timo/code/#incrasrfilter . An example configuration for the HelloNGram demo distributed with Sphinx-4 as well as a readme file are available. Comments are very welcome.
Cheers,
Timo
Timo Baumann, Michaela Atterer, and David Schlangen (2009).
"Assessing and Improving the Performance of Speech Recognition for Incremental Systems".
in Proceedings of NAACL-HLT 2009, Boulder, USA.
Hi,
I'm pleased to announce that the methods described in our recent NAACL paper [1] to filter intermediate ASR results and generate incremental ASR output are now implemented as a Sphinx-4 ResultListener. The code is available at http://www.ling.uni-potsdam.de/~timo/code/#incrasrfilter . An example configuration for the HelloNGram demo distributed with Sphinx-4 as well as a readme file are available. Comments are very welcome.
Cheers,
Timo
Timo Baumann, Michaela Atterer, and David Schlangen (2009).
"Assessing and Improving the Performance of Speech Recognition for Incremental Systems".
in Proceedings of NAACL-HLT 2009, Boulder, USA.
http://www.aclweb.org/anthology/N/N09/N09-1043.pdf
Great. Thanks Timo. I'll try this as soon as I can :)
antoine
Just an FYI. I'm using Sphinx4 out of the trunk. I grabbed it early Nov 2009.
I had to modify
CurrentHypothesis statusChanged
to
public void statusChanged(Recognizer.State status)
It now takes edu.cmu.sphinx.recognizer.Recognizer.State rather than State.
Good luck.