I want to know the best approaches to handle named entities (names of locations) when using automatic speech recognition systems (DeepSpeech2 for example).
So we want to see if there are any possible approaches to jointly consider named entity recognition when training the acoustic model or even correcting the output or giving some sort of priority to named entities.
- Badges
- Science topic
More Riadh Belkebir's questions See All
Similar questions and discussions