IPS-LMU / emuR

The main R package for the EMU Speech Database Management System (EMU-SDMS)
http://ips-lmu.github.io/EMU.html
23 stars 15 forks source link

Feature: faster runBASwebservice_all() #251

Open samgregory opened 3 years ago

samgregory commented 3 years ago

Hi @raphywink I have been using the runBASwebservice_all() recently on a relatively large database (it has a large number of short bundles) and noticed that this makes things considerably slower than if I used the MAUS web interface. The reason is because of the number of calls to the service to complete each bundle - once for tokenisation, once for pronunciation, once for force alignment and twice for pho2syl.

runBASwebservice_all() appears to be replicating the functionality of the Pipeline service - I'm wondering if it could be enhanced so it makes a call to Pipeline (providing the transcription and the audio) to do G2P->(CHUNKER)->MAUS->PHO2SYL and then converts the output?

I know the runBASwebservice_* was mostly worked on by @NPoe and that this feature would be fairly big change - what are your thoughts?