I have been trying citrinet model with a language model to improve the decoding for certain words. But with the use of the language model, I am finding that citrinet is skipping some words completely. For example, if I speak “I want to open a bank account”, the output is “I want to account”.
I am pasting a reference link here of another person having similar issues here
What is the solution for this problem? Any suggestions on what can be done here?