AL-BLEU is the Arabic Language MT automatic evaluator, a modified version of BLEU.
It uses MADA analysis of the translation and reference. 

Usage:    
python albleu.py <path-to-evaluation-directory>
    
Input arguments:
<path-to-evaluation-directory>: must contain: e-hyp (tranlsation), e-test (reference), e-hyp.mada, and e-test.mada
    
Note: MADA analysis files should be created specifically for the e-hyp/e-test files, i.e. every line in e-hyp corresponds to a line in e-hyp.mada in the same order. Same thing goes for e-test.