args = p.parse_args()
last_n = None
if args.gold_aln:
gold_alns = load_aln(args.gold_aln)
last_n = len(gold_alns)
e_snts = load_snts(args.e, args.limit)
f_snts = load_snts(args.f, args.limit)
assert len(e_snts) == len(f_snts)
gold_e_snts = load_snts(args.gold_e)
gold_f_snts = load_snts(args.gold_f)
ga = GizaAligner()
test_alns = ga.temp_train(e_snts + gold_e_snts, f_snts + gold_f_snts)
if args.gold_aln:
assert len(e_snts + gold_e_snts) >= len(gold_alns)
# -------------------------------------------
# We will assume that the test sentences are the last
# in the file.
# -------------------------------------------
alns_to_test = test_alns[-len(gold_alns) :]
assert len(alns_to_test) == len(gold_alns)
ae = AlignEval(alns_to_test, gold_alns)
print(",".join([str(args.limit)] + ["{:.5f}".format(f) for f in ae.all()]))
def train_aligner(o, e, f):
ga = GizaAligner()
# ga = GizaAligner.load(o, e, f)
ga.train(o, e, f)