I would like to hire someone to set up Sphinx-4 for me on a Linux system.
I want to be able to submit a list of wav (or mp3, if possible) files to Sphinx-4 and generate corresponding text transcripts of the recordings (.txt files named according to the original audio filename).
The audio recordings would be from a large number of speakers, so speaker-dependent training could not be done.
I would like to set this up with the 64,000 word HUB4 model file.
The process should be able to be activated by running a curl command on the server with a parameter for 1) the list of audio files and 2) the output directory.
I expect whoever accepts this job to have used Sphinx4 before or have extensive Java+Linux experience. Please show/explain your expertise in this area. I'd like to have more Sphinx4 customizations done after this task is completed (confidence ratings on words/sentences, etc).