Way to see Diarization Speaker-Probabilities by word?

[gatestonjohns]

Essentially, I am simply looking for a way to see the diarization probabilities for each speaker on any given word (or segment, if word is not possible). I have tracked down the functionality to be within the apply function of the speaker_diarization pipeline, but it is difficult to understand what is going on exactly without serious examination and time commitment. Any help would be appreciated, thanks!

[hbredin]

Related to #1636. Duplicating my answer here:

Out of the box, there is no such thing available in pyannote.

This is a tricky problem because the underlying algorithm is a combination of multiple parts, each of them having their own uncertainties, and some of them (e.g. the clustering part) don't even provide such a thing.

[gatestonjohns]

Is there any way we may be able to calculate a rough estimate of the diarization confidence using the soft_clusters matrix returned by the Agglomerative Clustering algorithm? I do not need an exactly correct probability matrix, just some type of indication when words are not super confidently diarized.

[qalabeabbas49]

Hi, any progress on this? I am also looking for something similar.

Thanks