#!/usr/bin/env python

# this script scores your recognition results. run it like this:
#
#  python score.py ref.txt hyp.txt
#
# the first file is the reference; the second file is the hypothesis.
#

# import required modules:
#  note that the path to the module htkmfc must be included in the
#  PYTHONPATH environment variable.
#
import os
import sys
import numpy as np

# main: this is the main function of this Python
#
def main(argv):

    # load the two files
    #
    refs = [line.rstrip('\n') for line in open(sys.argv[1])]
    hyps = [line.rstrip('\n') for line in open(sys.argv[2])]

    if len(refs) != len(hyps):
        print("error: files are not the same length")
        exit(0)

    # create a 3x3 array for scoring
    #
    conf = np.zeros((3, 3))

    # loop over the lists and check the first element
    #
    for r,h in zip(refs, hyps):
        ref = r.split(',')[0]
        hyp = h.split(',')[0]

        # count errors
        #
        conf[int(ref), int(hyp)] += int(1)

    # dump the confusion matrix
    #
    print("%2s  %6s %6s %6s" % ("", "c0", "c1", "c2"))
    print("%2s: %6d %6d %6d" % ("c0", conf[0,0], conf[0,1], conf[0,2]))
    print("%2s: %6d %6d %6d" % ("c1", conf[1,0], conf[1,1], conf[1,2]))
    print("%2s: %6d %6d %6d" % ("c2", conf[2,0], conf[2,1], conf[2,2]))

    # summarize the errors
    #
    print("error rate  = %10.4f%%" %
          ((1 - (conf[0,0] + conf[1,1] + conf[2,2])/len(refs)) * 100))

# begin gracefully
#
if __name__ == "__main__":
    main(sys.argv[0:])

#
# end of file