So first of all: a big thanks to the developers for this amazing open source project
After downloading and testing the pretraind model for deepspeech 0.6.0 i am pleased with the results but i would like to add words to the vocabulary (the trie/ language model).
My question is: can i further train the language model/trie that is already provided and if so, how can i do this? (And maybe what kind of changes could i expect performance wise?)
If not, which text resources would be recommended to train this using kenLM
ps: im also kind of new to language models so can anyone explain me how the alpha and beta weights work? And what the influence is in changing them?