def test_accuracy():
graphs_fname = config["test_data_dir"] + "/graphs_sample_out_de-en.pkl"
graphs = cPickle.load(open(graphs_fname))
ref_fname = config["test_data_dir"] + "/lemma_sample_out_de-en.ref"
result = accuracy_score(graphs[:1], ref_fname, "freq_score")
assert result.correct == 5
assert result.incorrect == 1
assert result.ignored == 1
assert_almost_equal(result.score, 0.8333333)
("incorrect", "i"),
("ignored", "i"),
("accuracy", "f"),
("graphs", "i"),
("nist", "f"),
("bleu", "f"),
("exp_name", "S128"),
]
new_results = np.zeros(len(old_results), descriptor)
for i, exp in enumerate(old_results):
ref_fname = config["eval"][exp["data"]][exp["source"] + "-" + exp["target"]]["lemma_ref_fname"]
graphs_fname = "_{}/{}_graphs.pkl".format(name, exp["exp_name"])
graphs = cPickle.load(open(graphs_fname))
accuracy = accuracy_score(graphs, ref_fname, name + "_score")
new_results[i]["graphs"] = len(graphs)
new_results[i]["data"] = exp["data"]
new_results[i]["source"] = exp["source"]
new_results[i]["target"] = exp["target"]
new_results[i]["min_count"] = exp["min_count"]
new_results[i]["max_freq"] = exp["max_freq"]
new_results[i]["correct"] = accuracy.correct
new_results[i]["incorrect"] = accuracy.incorrect
new_results[i]["accuracy"] = accuracy.score
new_results[i]["ignored"] = accuracy.ignored
new_results[i]["nist"] = exp["nist"]
new_results[i]["bleu"] = exp["bleu"]
new_results[i]["exp_name"] = exp["exp_name"]
np.save("_" + name + "-acc.npy", new_results)
