I have seen this being discussed earlier but not in much detail.
The question pertained to using DeepSpeech model to recognise names from, say, a database. Class-based LM was brought up but those threads seem to be abandoned without anyone actually posting about whether or not they successfully implemented such a solution. Has anyone tried using that and could perhaps provide guidelines on how to go about it?
Plus, would using a class-based LM mean that you can’t use DeepSpeech for general Speech-to-Text?