Would be nice to be able to choose which column in the CSV to compare (beside of speaker_type)
Would be good to allow custom mappings for these columns: e.g. I may want to map original speaker types to a small set, which could be useful for the LENA. It seems that all segments which are not MAL/FEM/CHI/OCH are treated as 'silence', but the LENA has other speech segments that do not belong to silence (e.g.: MAF/FAF)