#! /usr/bin/env python

""" A simple metric to compare two translation strings.

    USAGE: ./compare.py <human translation> <machine translation>
    
    Elias Zeidan | GPL | April 2012
"""
import sys
from nltk.tokenize import *

def add_filler(sent1, sent2):
        difference = len(sent1) - len(sent2)
        for i in range(difference):  
            sent2.append("blah")         
        return sent2     


def compare(human, machine):
    human_token = word_tokenize(human)
    machine_token = word_tokenize(machine)
    # print human_token, machine_token
    total_unmatched = 0     
    if len(human_token) != len(machine_token):
	print len(human_token), len(machine_token)	
	if len(human_token) > len(machine_token):
        	machine_token = add_filler(human_token, machine_token)     
    	else:
        	human_token = add_filler(human_token, machine_token)
    # print human, machine
    # print human_token, machine_token
    for i in range(min(len(human_token), len(machine_token))):         
        if human_token[i] != machine_token[i]:             
            total_unmatched += 1
    print "human translation: %s" % human
    print "machine translation: %s" % machine     
    return float(total_unmatched)/len(human_token)

def main():
    print compare(human, machine)

if __name__ == "__main__":
    human = sys.argv[1]
    machine = sys.argv[2]
    main()