Closed cqx820 closed 6 years ago
Take a look at our model adaptation script, which supports adapting a model to a custom dictionary and language model:
In the https://github.com/gooofy/zamia-speech#model-adaptation you mentioned, and https://github.com/gooofy/kaldi-adapt-lm there seemed only limited to some existing vocabulary. Is there some configuration that can use my own vocabulary as well as LM? thanks
Also, have a trouble understanding the sampa format, current kaldi seemed only use CMU phoneset . Is there a way that I can use the kaldi trained g2p model (trained with CMU phoneset)?
speech_kaldi_adapt.py support custom dicts as well as a custom language model:
$ ./speech_kaldi_adapt.py -h Usage: speech_kaldi_adapt.py [options] src_model_dir dict (lm.arpa|G.src.fst|grammar.jsgf) dst_model
Options: -h, --help show this help message and exit -v, --verbose enable verbose logging
Hi,
I would like adding a new word, but this word doesn't exist in CMU pronounce dictionary. So I used logios lexicon tool to generate a new pronunciation dictionary. I would like to know how to add this new dictionary to the existing model? Thank you so much.