def run(output_dir, feature_num, mode):
feature_num = int(feature_num)
taxo_res = TaxonomyResources(freq_fpaths, isa_fpaths)
taxo_features = TaxonomyFeatures(taxo_res, relations_fpath=train_relations_fpath)
ensure_dir(output_dir)
features = ["hyper_in_hypo_i","hypo2hyper_substract", "freq_substract", "in_weight_substract", "length_substract",
"hypo2hyper_s_substract","hypo2hyper_max2_substract"]
features = features[:feature_num]
if mode == "gridsearch":
# grid search is only supported for SVC
method = "SVC"
hc = SuperTaxi(join(output_dir, "SVC-grid-search"), method="SVC", features=features, overwrite=True)
clf = hc.grid_search_svc(taxo_features.relations, test=TEST_GRIDSEARCH)
return
for method in METHODS:
try:
classifier_dir = join(output_dir, method)
print("\n", method.upper(), "\n", "="*50)
hc = SuperTaxi(classifier_dir, method=method, features=features, overwrite=True)
if mode == "train":
clf = hc.train(taxo_features.relations)
hc._print_clf_info()
elif mode == "cv":
hc.crossval(taxo_features.relations)
else:
print("Error: unrecognised mode %s" % mode)
except:
print(format_exc())
def evaluate_on_trial_taxo():
relations_fpath = join(
RES_DIR, "relations.csv"
) # assuming features "hyper_in_hypo_i" and "hypo2hyper_substract"
taxo_fpath = relations_fpath + "-taxo.csv"
print("Relations:", relations_fpath)
print("Unpruned taxonomy:", taxo_fpath)
taxo_features = TaxonomyFeatures(TaxonomyResources(),
relations_fpath=relations_fpath,
lang="en")
taxo_predict = TaxonomyPredictor(taxo_features)
taxo_predict.predict_by_global_threshold(threshold=0,
field="hypo2hyper_substract",
or_correct_predict=False)
taxo_predict.predict_by_global_threshold(threshold=0,
field="hyper_in_hypo_i",
or_correct_predict=True)
taxo_predict.save(taxo_fpath)
taxo_predict.evaluate(field="correct_predict")
for max_knn in [1, 2, 3, 5]:
taxo_knn_fpath = relations_fpath + "-taxo-knn" + str(max_knn) + ".csv"
taxo_predict.predict_by_local_threshold(threshold=0,
max_knn=max_knn,
field="hypo2hyper_substract",
or_correct_predict=False)
taxo_predict.predict_by_global_threshold(threshold=0,
field="hyper_in_hypo_i",
or_correct_predict=True)
taxo_predict.save(taxo_knn_fpath)
taxo_predict.evaluate(field="correct_predict")
def extract_semeval_taxo(input_voc_pattern, language, mode, classifiers_pattern, test_en):
#Laedt alle Datensaetze(auch alle Domaenen, aus vocabularies)
taxo_res_common, taxo_res_domain = load_res(language, mode, test_en)
for voc_fpath in sorted(glob(input_voc_pattern)):
for space in [False, True]:
s = "-space" if space else ""
relations_fpath = voc_fpath + s + "-relations.csv"
taxo_fpath = relations_fpath + "-taxo.csv"
print "\n", voc_fpath, "\n", "="*50
print "Relations:", relations_fpath
print "Unpruned taxonomy:", taxo_fpath
#Laedt domain-datenset und kombiniert sie mit dem allgemeinen Datenset
taxo_res_domain_voc = get_taxo_res_domain_voc(taxo_res_domain, voc_fpath)
taxo_res_voc = combine_taxo_res(taxo_res_common, taxo_res_domain_voc)
taxo_features = TaxonomyFeatures(taxo_res_voc, voc_fpath, lang=language)
if mode == "simple":
taxo_features.fill_direct_isas()
taxo_features.fill_substrings(must_have_space=space)
taxo_features.hypo2hyper_ratio()
taxo_predict = TaxonomyPredictor(taxo_features)
taxo_predict.predict_by_global_threshold(threshold=0, field="hypo2hyper_substract", or_correct_predict=False)
taxo_predict.predict_by_global_threshold(threshold=0, field="hyper_in_hypo_i", or_correct_predict=True)
taxo_predict.save(taxo_fpath)
for max_knn in [1, 2, 3, 5]:
#hypo2hyper fuer pattern
#hyperinhypoi feur substring
taxo_knn_fpath = relations_fpath + "-taxo-knn" + unicode(max_knn) + ".csv"
taxo_predict.predict_by_local_threshold(threshold=0, max_knn=max_knn, field="hypo2hyper_substract", or_correct_predict=False)
taxo_predict.predict_by_global_threshold(threshold=0, field="hyper_in_hypo_i", or_correct_predict=True)
taxo_predict.save(taxo_knn_fpath)
elif mode == "super":
taxo_features.fill_super_features()
for classifier_dir in glob(classifiers_pattern):
try:
print "Predicting with:", classifier_dir
taxo_predict = TaxonomyPredictor(taxo_features)
method = taxo_predict.predict_by_classifier(classifier_dir)
taxo_predict.save(taxo_fpath + "-" + method + ".csv")
taxo_predict.save(taxo_fpath + "-" + method + "-conf.csv", conf=True)
except:
print format_exc()
def extract_semeval_taxo(input_voc_pattern, language, mode, classifiers_pattern):
taxo_res_common, taxo_res_domain = load_res(language, mode)
for voc_fpath in sorted(glob(input_voc_pattern)):
for space in [False]: #, True]:
s = "-space" if space else ""
relations_fpath = voc_fpath + s + "-relations.csv"
taxo_fpath = relations_fpath + "-taxo.csv"
print "\n", voc_fpath, "\n", "="*50
print "Relations:", relations_fpath
print "Unpruned taxonomy:", taxo_fpath
taxo_res_domain_voc = get_taxo_res_domain_voc(taxo_res_domain, voc_fpath)
taxo_res_voc = combine_taxo_res(taxo_res_common, taxo_res_domain_voc)
taxo_features = TaxonomyFeatures(taxo_res_voc, voc_fpath, lang=language)
if mode == "simple":
taxo_features.fill_direct_isas()
taxo_features.fill_substrings(must_have_space=space)
taxo_features.hypo2hyper_ratio()
taxo_predict = TaxonomyPredictor(taxo_features)
taxo_predict.predict_by_global_threshold(threshold=0, field="hypo2hyper_substract", or_correct_predict=False)
taxo_predict.predict_by_global_threshold(threshold=0, field="hyper_in_hypo_i", or_correct_predict=True)
taxo_predict.save(taxo_fpath)
for max_knn in [1, 2, 3, 5]:
taxo_knn_fpath = relations_fpath + "-taxo-knn" + unicode(max_knn) + ".csv"
taxo_predict.predict_by_local_threshold(threshold=0, max_knn=max_knn, field="hypo2hyper_substract", or_correct_predict=False)
taxo_predict.predict_by_global_threshold(threshold=0, field="hyper_in_hypo_i", or_correct_predict=True)
taxo_predict.save(taxo_knn_fpath)
elif mode == "super":
taxo_features.fill_super_features()
for classifier_dir in glob(classifiers_pattern):
try:
print "Predicting with:", classifier_dir
taxo_predict = TaxonomyPredictor(taxo_features)
method = taxo_predict.predict_by_classifier(classifier_dir)
taxo_predict.save(taxo_fpath + "-" + method + ".csv")
taxo_predict.save(taxo_fpath + "-" + method + "-conf.csv", conf=True)
except:
print format_exc()
from jnt.isas.taxo import TaxonomyResources, TaxonomyFeatures relations_fpath = "/Users/alex/tmp/semeval/new/relations.csv" voc_fpath = "/Users/alex/tmp/semeval/new/voc.csv" isa_fpath = "/Users/alex/tmp/semeval/new/en_dt.csv-isas.csv" taxo_res = TaxonomyResources(freq_fpaths=[""], isa_fpaths=[isa_fpath]) taxo_features = TaxonomyFeatures(taxo_res, voc_fpath=voc_fpath, lang='en') taxo_features.fill_direct_isas()
