wav2vec2-sprint
wav2vec2-sprint copied to clipboard
Generate the N-best (top few) hypotheses
Hello Everyone,
I'm currently working on a project involving Wav2vec2 Automatic Speech Recognition (ASR) system. Specifically, I'm trying to generate an N-best list of hypotheses using the Wav2vec2 ASR.
I've made some progress, but I'm encountering a few challenges with the code implementation. I was wondering if anyone in the community has experience with this and could provide some guidance or code snippets.
My goal is to generate the N-best hypotheses along with their confidence scores.
Thanking You