search

JoFrhwld / FAVE

A repository for maintaing the fave-align and fave-extract toolkits
GNU General Public License v3.0
114 stars 37 forks source link

Feature/parselmouth fix #85

Open JoFrhwld opened 1 year ago

JoFrhwld commented 1 year ago

All good points. Re: speech software deprecation, I'm not sure we need to be worried about dropping esps support since, as far as I can tell, it hasn't been actually acquirable software for about a decade or more (all links to it 404).

But, I think retaining checkSpeechSoftware, perhaps using parselmouth as a fallback or its own specifiable option is a good idea. Without wholesale deprecating the praat option & adding parselmouth as an option, this could probably still make it into 2.1? All that writing to disc is a real blocker to any kind of parallelization, which is why I'm eager to firmly set us down this road.

chrisbrickhouse commented 1 year ago

All good points. Re: speech software deprecation, I'm not sure we need to be worried about dropping esps support since, as far as I can tell, it hasn't been actually acquirable software for about a decade or more (all links to it 404).

Eh, I guess I can get behind removal. HTK's also been moribund for a while which is where my hesitation comes from, but that's at least kinda available---ESPS isn't. As long as we're super clear about the change in the release notes, I think you're right that this is low risk.

But, I think retaining checkSpeechSoftware, perhaps using parselmouth as a fallback or its own specifiable option is a good idea. Without wholesale deprecating the praat option & adding parselmouth as an option, this could probably still make it into 2.1?

I've added some deprecation warnings in 83cbde6 to that effect. Could be specified more (and probably include a @versionchanged decorator).

All that writing to disc is a real blocker to any kind of parallelization, which is why I'm eager to firmly set us down this road.

Hmm, that feature wasn't clear to me from the PR. I agree that's a priority and should be included in 2.1. I think the combo of deprecation, esps removal, and parselmouth inclusion is a good one. The subclassing and such can be documented as technical debt and fixed later.

JoFrhwld commented 1 year ago

I haven't really messed around with the Github Project kanban boards since I set them up, but could be useful for planning a little bit of this out. https://github.com/users/JoFrhwld/projects/1/views/1

What I've managed to figure out about the history of why ESPS was included at all is from here: http://www.phon.ox.ac.uk/releases Specifically

Subsequently, Stephen Isard (Institute for Research in Cognitive Science, University of Pennsylvania) updated the sources to compile and run on current OSX and Linux machines, and Danny Yee (University of Oxford) has packaged them as a .deb file for Ubuntu 12.04. The licence terms are exactly those of the final ESPS source code release, without modification.

Stephan Isard was a co-author with Keelan Evanini and Mark Liberman on the first FAVE-like system in 2008, and Ubuntu 12.04 was initially released in 2012, all immediately prior to the first verisons of the FAVE-suite being created.