speech recognition roadmap

An overview of the current state and way forward for speech recognition on Linux. Implementation details will appear in further specs.

2007-05-14 kamion: Not sure if this makes sense as a gutsy goal yet; proposing it so that we can consider further.

2007-05-19 heno: agree, it's not a gutsy goal, more of a rolling start. It was useful to have sessions on it at UDS though. There is a Gnome SoC project in this area, so we should see some progress.

2008-11-15: The SoC project (gnome voice control) was completed, doesn't work in the slightest, and has apparently been abandoned since then.

2008-11-15: RainCT: I just want to note that I packaged julius and VoxForge (julius-voxforge) and that both are in Intrepid.

2010-06-18: pcgaldo: In addition to a recognition engine, it is necessary for the end user to have an intuitive interface to modify the settings. Something like "Simon Listens" would be valid as a starting point: http://simon-listens.org/
It is also necessary to facilitate the incorporation of models to VoxForge free speech corpus. Having a step by step wizard, with GUI, to carry out recordings to incorporate models, would be the best way.

2011-02-21: I'm working on a project (https://launchpad.net/cmusphinx-train) that is quite similar to the description the pcgaldo had in the second sentence. It'll be able to adapt ARPA voice models of which can be used with CMU Sphinx and Julius. An alpha with these features will be available some time in 2011-03.