speak content grounding?

[MonolithFoundation]

Any way to do speak content grounding? is like check a sentence is presented in audio or not (not caption

[wsntxxn]

Sorry current checkpoints do not support speech content grounding, which is more related to keyword spotting. However, due to the similarity in task definition, I think the same pipeline can be applied on speech content grounding. My consideration is that with the frame-level annotation from HMM models, the weakly-supervised method is not necessary. The whole model can be trained using frame-level labels.