Decoding using LM with Contextual biasing (Hotwords)

[kerolos]

1- Is it possible to use a LM (RNNLM or n-gram) at the same time adding Hotwords in ONNX ?
2- How can I change the display word, for example word contains special characters or with hyphen. I had tried to put the desired special words with **@**SPECIALWORD in ( --hotwords-file ) , but it does not work (the tokens_type i used is bpe) ?, However in the old system By using Phone Based lexicon we can display any words have special character !
Thanks in advance