analyze its to see how many child vocs there are in total
if possible, pull out 250 randomly & 250 from high child volubility 1-minute regions (the latter should be "consecutive vocs" ie take top minute and pull out all the vocs from there, then move to next minute, etc)
just realized that may be problematic: what if you find the same vocs across both methods? in that case, I think we should pull out the segment only once, but keep the label so that it's analyzed in both groups (when we do anas)