Adding new words to already trained model

[ChrystianKacki]

I have trained a zipformer transducer streaming model using Icefall with quite a good WER of 2.04%. I have exported it and I am using it now with Sherpa-ONNX. Unfortunately there are some difficult words which where not present in training dataset and they are not recognized properly. Can I use hotwords function of Sherpa-ONNX to introduce new words ? What value of hotwords_score should I set in model config ? Can hotwords file contain many thousands of words ?