Review of 2018-04-04am-KCH

[aarppe]

For this session, before reuploading the snippets consisted mainly of noise.

After the reannotation, there are many good snippets, but sometimes these appear to come from another session, e.g. 2018-03-14am-KCH, rather than from 2018-04-04am-KCH. And actually, alongside the good snippets from another session, effectively all audio snippets from 2018-04-04am-KCH are bad. Examples can be found e.g. on the following validation page.

@fbanados Are these bad snippets the remnants of the earlier incorrect annotation?

Otherwise, the exclude feature has now proven most useful in leaving out the bad audio snippets, that do not contain anything of value.

[fbanados]

Yes that would be my assumption. Good the exclude feature is helpful.

Because it is computationally difficult to declare a sound file as noise, it would be now hard to distinguish those from others. Mostly because the simplest approach is to first delete all recordings from a session and then reupload it (which I clearly didn't do here), but that would remove some recordings that are already annotated as best. The alternatives are either to delete all recordings marked as excluded from the session or re-import it again (and the reimported sound files should be identified as already present) and delete all those not marked for reimporting (even if not replaced... that would keep the best annotations)

I could eventually run a script for either of those approaches if that would be preferred.

[aarppe]

As discussed, we could filter out the snippets marked as noise, basing that on those snippets marked for exclusion (and being bad), after standardization and validation for this particular session has been completed, in a week or two.

[fbanados]

Let me know when this is ready to be done.