Tue Nov 12 14:51:57 EST 1996 msiegler
Clusters utterances together based on Symmetric Cross Entropy values
also known as the KL2 metric, derived from single-mode distrbutions
on the .mfc files,

Uses agglomerative clustering technique to try to build clusters meeting the
criterion that the distance between items in a cluster is less than [-t] and
the total amount of frames in each cluster is at least [-m]

Usage: UTT_cluster
-i <ctl_file> -d <mfc_dir> -e <mfc_ext> 
-r <report_file> [-t (distance threshold (0-1) def=0.00)]
[-m minimum cluster size in frames def=0 (none)] [-v (verbose)]


control file format looks like

<.mfc file source> <start frame> <end frame> <utterance id>
The report file takes the following form:

<utterance id> <cluster #>



For the CMU Hub 4 1996 evaluation we used:
	-t 0.025 -m 1000

