sherpa-onnx icon indicating copy to clipboard operation
sherpa-onnx copied to clipboard

Published 20 hours ago • verified publisher icon k2-fsa

Adding new words using Hotwords option with multiple variants per word.

Open kerolos opened this issue 1 year ago • 0 comments
trafficstars

I have encountered an issue with the word PRECENT not being recognized. I attempted to generate a BPE for this word using the tool sherpa-onnx/scripts/text2token.py, which provided only one variant: ▁PER C ENT. However, it is not recognized in most cases.

Is there any way to produce more variants per word to ensure recognition, or should I adjust parameters such as boosting score, trigger threshold for this word, or increase the hotwords_score?

kerolos avatar Apr 18 '24 08:04 kerolos