Open renepickhardt opened 10 years ago
In our experimentation we found that we won't be able to conform with standard formats (namely APRA) I guess we won't have this for stable release then?
yes our current tool will not provide an arpa file since a good decoder is shipped. I would still leave this bug open (maybe move it to a different milestone) since saving our model as an FST might still be an option also for saving space and improving runtime
the output of language models and n-grams should follow standard formats e.g.
weighted finite state transducer format (WFST) ARPA format
currently I am not sure if more formats exist. they should be researched and implemented.