def round_robin(self, config_path_learning, config_path, feature_set, lps):
config = ConfigParser()
config.readfp(open(config_path))
with open(config_path_learning, "r") as cfg_file:
config_learning = yaml.load(cfg_file.read())
f_results = open("results.txt", "w")
for test_lp in sorted(lps):
x_train = os.path.expanduser(self.config.get('Data', 'output_dir')) + "/" + "x_" + self.config.get("Settings", "dataset") + "." + "train" + "." + "tsv"
y_train = os.path.expanduser(self.config.get('Data', 'output_dir')) + "/" + "y_" + self.config.get("Settings", "dataset") + "." + "train" + "." + "tsv"
x_test = os.path.expanduser(self.config.get('Data', 'output_dir')) + "/" + "x_" + self.config.get("Settings", "dataset") + "." + "test" + "." + "tsv"
y_test = os.path.expanduser(self.config.get('Data', 'output_dir')) + "/" + "y_" + self.config.get("Settings", "dataset") + "." + "test" + "." + "tsv"
train_lps = ScoringTask.get_train_lps(lps, test_lp)
train_feature_values = []
train_reference_values = []
test_feature_values = read_features_file(os.path.expanduser(self.config.get('Data', 'output_dir')) + "/" + \
"x_" + self.config.get("Settings", "dataset") + "." + feature_set + "." + test_lp + "." + "tsv", "\t")
test_reference_values = read_reference_file(os.path.expanduser(self.config.get('Data', 'output_dir')) + "/" + \
"y_" + self.config.get("Settings", "dataset") + "." + feature_set + "." + test_lp + "." + "tsv", "\t")
for train_lp in sorted(train_lps):
feature_values = read_features_file(os.path.expanduser(self.config.get('Data', 'output_dir')) + "/" + \
"x_" + self.config.get("Settings", "dataset") + "." + feature_set + "." + train_lp + "." + "tsv", "\t")
reference_values = read_reference_file(os.path.expanduser(self.config.get('Data', 'output_dir')) + "/" + \
"y_" + self.config.get("Settings", "dataset") + "." + feature_set + "." + train_lp + "." + "tsv", "\t")
train_feature_values += list(feature_values)
train_reference_values += list(reference_values)
write_feature_file(x_train, train_feature_values)
write_reference_file(y_train, train_reference_values)
write_feature_file(x_test, test_feature_values)
write_reference_file(y_test, test_reference_values)
gold_standard = test_reference_values
predictions = ScoringTask.train_predict(config_path_learning)
# predictions = ScoringTask.recursive_feature_elimination(config_learning, config, 50)
correlation = ScoringTask.evaluate_predicted(predictions, gold_standard)
f_results.write(test_lp + " " + str(correlation) + " with " + feature_set + "\n")
os.remove(x_train)
os.remove(x_test)
os.remove(y_train)
os.remove(y_test)
def recursive_feature_elimination(config_learning, config_data, number_features):
output = open(os.path.expanduser(config_data.get("Learner", "models")) + "/" + "feature_ranks.txt", "w")
feature_names = FeatureExtractor.get_combinations_from_config_file_unsorted(config_data)
x_train = read_features_file(config_learning.get('x_train'), '\t')
y_train = read_reference_file(config_learning.get('y_train'), '\t')
x_test = read_features_file(config_learning.get('x_test'), '\t')
estimator, scorers = learn_model.set_learning_method(config_learning, x_train, y_train)
scale = config_learning.get("scale", True)
if scale:
x_train, x_test = scale_datasets(x_train, x_test)
rfe = RFE(estimator, number_features, step=1)
rfe.fit(x_train, y_train)
for i, name in enumerate(feature_names):
output.write(name + "\t" + str(rfe.ranking_[i]) + "\n")
print(name + "\t" + str(rfe.ranking_[i]))
predictions = rfe.predict(x_test)
output.close()
return predictions
def training_set_for_learn_to_rank_from_feature_file(config_learning, config):
data_set_name = config.get('WMT', 'dataset')
feature_values = read_features_file(config_learning.get("x_train", None), '\t')
human_rankings = read_reference_file(config_learning.get("y_train", None), '\t')
new_features = []
new_labels = []
path_features = os.path.expanduser(config.get('WMT', 'output_dir')) + '/' + 'x_' + data_set_name + ".learn_rank.tsv"
path_objective = os.path.expanduser(config.get('WMT', 'output_dir')) + '/' + 'y_' + data_set_name + ".learn_rank.tsv"
for i, label in enumerate(human_rankings):
if label == 2:
features = np.subtract(RankingTask.split_list(feature_values[i])[0], RankingTask.split_list(feature_values[i])[1])
new_label = 1
else:
features = np.subtract(RankingTask.split_list(feature_values[i])[1], RankingTask.split_list(feature_values[i])[0])
new_label = 0
new_features.append(features)
new_labels.append(new_label)
write_feature_file(path_objective, new_features)
write_reference_file(path_features, new_labels)
def open_datasets(train_path, train_ref_path, test_path,
test_ref_path, delim, labels_path=None, tostring=False):
if not os.path.isfile(os.path.abspath(train_path)):
raise IOError("training dataset path is not valid: %s" % train_path)
if not os.path.isfile(os.path.abspath(train_ref_path)):
raise IOError("training references path is not valid: %s" % train_ref_path)
if not os.path.isfile(os.path.abspath(test_path)):
raise IOError("test dataset path is not valid: %s" % test_path)
if not os.path.isfile(os.path.abspath(test_ref_path)):
raise IOError("test references path is not valid: %s" % test_ref_path)
labels = []
if labels_path is not None:
if not os.path.isfile(os.path.abspath(labels_path)):
raise IOError("labels file is not valid: %s" % labels_path)
labels = read_labels_file(labels_path, delim)
X_train = read_features_file(train_path, delim, tostring=tostring)
y_train = read_reference_file(train_ref_path, delim, tostring=tostring)
X_test = read_features_file(test_path, delim, tostring=tostring)
y_test = read_reference_file(test_ref_path, delim, tostring=tostring)
if len(X_train.shape) != 2:
raise IOError("the training dataset must be in the format of a matrix with M lines and N columns.")
if len(X_test.shape) != 2:
raise IOError("the test dataset must be in the format of a matrix with M lines and N columns.")
if X_train.shape[0] != y_train.shape[0]:
print(X_train.shape[0], y_train.shape[0])
raise IOError("the number of instances in the train features file does not match the number of references given.")
if X_test.shape[0] != y_test.shape[0]:
raise IOError("the number of instances in the test features file does not match the number of references given.")
if X_train.shape[1] != X_test.shape[1]:
raise IOError("the number of features in train and test datasets is different.")
return X_train, y_train, X_test, y_test, labels
def load_predict(config_learning, config_data):
learning_config = config_learning.get("learning", None)
method_name = learning_config.get("method", None)
x_train = read_features_file(config_learning.get('x_train'), '\t')
y_train = read_reference_file(config_learning.get('y_train'), '\t')
x_test = read_features_file(config_learning.get('x_test'), '\t')
y_test = read_reference_file(config_learning.get('y_test'), '\t')
scale = config_learning.get("scale", True)
if scale:
x_train, x_test = scale_datasets(x_train, x_test)
estimator = joblib.load(os.path.expanduser(config_data.get("Learner", "models")) + "/" + method_name + ".pkl")
predictions = estimator.predict(x_test)
return predictions
def get_data(self):
human_scores = read_reference_file(os.path.expanduser(self.config.get('Data', 'human_scores')), '\t')
process = Process(self.config)
sents_tgt, sents_ref = process.run_processors()
extractor = FeatureExtractor(self.config)
features_to_extract = FeatureExtractor.read_feature_names(self.config)
extractor.extract_features(features_to_extract, sents_tgt, sents_ref)
return extractor.vals, human_scores
def train_model(cfg, model_path):
x_train = read_features_file(cfg.get('x_train'), '\t')
y_train = read_reference_file(cfg.get('y_train'), '\t')
x_test = read_features_file(cfg.get('x_test'), '\t')
scale = cfg.get("scale", True)
if scale:
x_train, x_test = scale_datasets(x_train, x_test)
estimator, scorers = learn_model.set_learning_method(cfg, x_train, y_train)
estimator.fit(x_train, y_train)
joblib.dump(estimator, model_path)
def train_save(config_learning, config_data):
learning_config = config_learning.get("learning", None)
method_name = learning_config.get("method", None)
x_train = read_features_file(config_learning.get('x_train'), '\t')
y_train = read_reference_file(config_learning.get('y_train'), '\t')
x_test = read_features_file(config_learning.get('x_test'), '\t')
scale = config_learning.get("scale", True)
if scale:
x_train, x_test = scale_datasets(x_train, x_test)
estimator, scorers = learn_model.set_learning_method(config_learning, x_train, y_train)
estimator.fit(x_train, y_train)
joblib.dump(estimator, os.path.expanduser(config_data.get('Learner', 'models')) + '/' + method_name + '.pkl')
def clean_dataset(config_learning, human_comparisons):
feature_values = read_features_file(config_learning.get('x_train'), '\t')
labels = read_reference_file(config_learning.get('y_train'), '\t')
new_feature_values = []
new_labels = []
human_comparisons = RankingTask.eliminate_ties(human_comparisons)
comparisons_untied_phrases = defaultdict(list)
comparisons_untied_signs = defaultdict(list)
deduplicated_phrases, deduplicated_signs = HumanRanking.deduplicate(human_comparisons)
for dataset, lang_pair in sorted(human_comparisons.keys()):
for comparison in human_comparisons[dataset, lang_pair]:
if comparison.sign == "=":
continue
else:
comparisons_untied_phrases[dataset, lang_pair].append([comparison.phrase, comparison.sys1, comparison.sys2])
comparisons_untied_signs[dataset, lang_pair].append(comparison.sign)
for dataset, lang_pair in sorted(human_comparisons.keys()):
for i, comparison in enumerate(comparisons_untied_phrases[dataset, lang_pair]):
features = feature_values[i]
label = labels[i]
if comparison in deduplicated_phrases[dataset, lang_pair]:
if deduplicated_signs[dataset, lang_pair][deduplicated_phrases[dataset, lang_pair].index(comparison)] is None:
continue
label = RankingTask.signs_to_labels(deduplicated_signs[dataset, lang_pair][deduplicated_phrases[dataset, lang_pair].index(comparison)])
new_feature_values.append(features)
new_labels.append(label)
write_feature_file(config_learning.get('x_train') + "." + "clean", new_feature_values)
write_reference_file(config_learning.get('y_train') + "." + "clean", new_labels)
def recursive_feature_elimination_cv(config_learning, config_data):
output = open(os.path.expanduser(config_data.get("Learner", "models")) + "/" + "feature_ranks.txt", "w")
feature_names = FeatureExtractor.get_features_from_config_file_unsorted(config_data)
combination_methods = FeatureExtractor.get_combinations_from_config_file_unsorted(config_data)
x_train = read_features_file(config_learning.get('x_train'), '\t')
y_train = read_reference_file(config_learning.get('y_train'), '\t')
x_test = read_features_file(config_learning.get('x_test'), '\t')
estimator, scorers = learn_model.set_learning_method(config_learning, x_train, y_train)
scale = config_learning.get("scale", True)
if scale:
x_train, x_test = scale_datasets(x_train, x_test)
rfecv = RFECV(estimator=estimator, step=1, cv=StratifiedKFold(y_train, 2), scoring='accuracy')
rfecv.fit(x_train, y_train)
feature_list = []
for i, feature_name in enumerate(feature_names):
if combination_methods[i] == 'both':
feature_list.append(feature_name)
feature_list.append(feature_name)
else:
feature_list.append(feature_name)
for i, name in enumerate(feature_list):
output.write(name + "\t" + str(rfecv.ranking_[i]) + "\n")
output.close()
predictions = rfecv.predict(x_test)
return predictions