def evaluate_simi(wv, w2i, vocab):
wv_dict = dict()
for w in vocab:
wv_dict[w] = wv[w2i[w], :]
if isinstance(wv_dict, dict):
w = Embedding.from_dict(wv_dict)
# Calculate results on similarity
print("Calculating similarity benchmarks")
similarity_tasks = {
"WS353": fetch_WS353(),
"RG65": fetch_RG65(),
# "WS353R": fetch_WS353(which="relatedness"),
# "WS353S": fetch_WS353(which="similarity"),
"SimLex999": fetch_SimLex999(),
"MTurk": fetch_MTurk(),
"RW": fetch_RW(),
"MEN": fetch_MEN(),
}
# similarity_results = {}
for name, data in iteritems(similarity_tasks):
print(
"Sample data from {}, num of samples: {} : pair \"{}\" and \"{}\" is assigned score {}"
.format(name, len(data.X), data.X[0][0], data.X[0][1], data.y[0]))
score = evaluate_similarity(w, data.X, data.y)
print("Spearman correlation of scores on {} {}".format(name, score))
def web_tests(emb):
"""
:param emb: dict of words and their corresponding embeddings
:return: dict of word-embeddings-benchmarks tests and scores received
"""
similarity_tasks = {
'WS353': fetch_WS353(),
'RG65': fetch_RG65(),
'RW': fetch_RW(),
'MTurk': fetch_MTurk(),
'MEN': fetch_MEN(),
'SimLex999': fetch_SimLex999()
}
web_emb = Embedding(Vocabulary(list(emb.keys())), list(emb.values()))
similarity_results = {}
for name, data in iteritems(similarity_tasks):
similarity_results[name] = evaluate_similarity(web_emb, data.X, data.y)
logging.info("Spearman correlation of scores on {} {}".format(
name, evaluate_similarity(web_emb, data.X, data.y)))
return similarity_results
def get_dataset(dataset_name):
if dataset_name == "WS353":
return fetch_WS353("similarity")
elif dataset_name == "MEN":
return fetch_MEN("all")
elif dataset_name == "SimLex-999":
return fetch_SimLex999()
elif dataset_name == "MTurk":
return fetch_MTurk()
elif dataset_name == "WS353":
return fetch_WS353('all')
elif dataset_name == "RG65":
return fetch_RG65()
elif dataset_name == "RW":
return fetch_RW()
elif dataset_name == "TR9856":
return fetch_TR9856()
elif dataset_name == "MSR":
return fetch_msr_analogy()
elif dataset_name == "Google":
return fetch_google_analogy()
else:
raise Exception("{}: dataset not supported".format(dataset_name))
def evaluateOnAll(w):
similarity_tasks = {
"MTurk": fetch_MTurk(),
"MEN": fetch_MEN(),
"WS353": fetch_WS353(),
"RubensteinAndGoodenough": fetch_RG65(),
"Rare Words": fetch_RW(),
"SIMLEX999": fetch_SimLex999(),
"TR9856": fetch_TR9856()
}
similarity_results = {}
for name, data in iteritems(similarity_tasks):
similarity_results[name] = evaluate_similarity(w, data.X, data.y)
print("Spearman correlation of scores on {} {}".format(name, similarity_results[name]))
# Calculate results on analogy
print("Calculating analogy benchmarks")
analogy_tasks = {
"Google": fetch_google_analogy(),
"MSR": fetch_msr_analogy()
}
analogy_results = {}
for name, data in iteritems(analogy_tasks):
analogy_results[name] = evaluate_analogy(w, data.X, data.y)
print("Analogy prediction accuracy on {} {}".format(name, analogy_results[name]))
analogy_results["SemEval2012_2"] = calAnswersonSemEval(w)['all']
print("Analogy prediction accuracy on {} {}".format("SemEval2012", analogy_results["SemEval2012_2"]))
analogy = pd.DataFrame([analogy_results])
sim = pd.DataFrame([similarity_results])
results = sim.join(analogy)
return results
plt.style.use('ggplot')
np.random.seed(0)
# Configure logging
logging.basicConfig(format='%(asctime)s %(levelname)s:%(message)s', level=logging.DEBUG, datefmt='%I:%M:%S')
# SETUP
datasets = [('SL999', fetch_SimLex999()),
('SL333', fetch_SimLex999(which='333')),
('SV3500-t', fetch_SimVerb3500(which='test')),
('WS353', fetch_WS353()),
#('WS353R', fetch_WS353(which='relatedness')),
#('RW', fetch_RW()),
('MEN-t', fetch_MEN(which='test')),
('SCWS', fetch_SCWS()),
('MTurk', fetch_MTurk())]
logging.info(fuel.config.data_path)
vocab_defs_fname = os.path.join(fuel.config.data_path[0], "vocab.txt")
logging.info("using vocab for definition {}".format(vocab_defs_fname))
# END SETUP
parser = argparse.ArgumentParser("Evaluate embeddings")
parser.add_argument("emb_filename", help="Location of embeddings")
parser.add_argument("emb_format", help="either 'glove', 'dict' or 'dict_poly'")
parser.add_argument("root_dicts",
help="dirname that contains all.json and test.json")
parser.add_argument("--lowercase", action='store_true')
parser.add_argument("--normalize", action='store_true')
parser.add_argument("--vocab_size", default=None, type=int,
help="vocab size (GloVe only)")
def test_MTurk_fetcher():
data = fetch_MTurk()
assert (len(data.y) == len(data.X) == 287)
assert (10.0 >= data.y.max() >= 9)
'CBOW': 'fetch_CBOW()'
}
#w_glove = fetch_GloVe(corpus="wiki-6B", dim=300)
#w_PDC = fetch_PDC()
#w_HDC = fetch_HDC()
#w_LexVec = fetch_LexVec(which="wikipedia+newscrawl-W")
#w_conceptnet_numberbatch = fetch_conceptnet_numberbatch()
#w_w2v = fetch_SG_GoogleNews()
#w_fastText = fetch_FastText() #load_embedding(path, format='word2vec', normalize=True, lower=False, clean_words=False)
# Define tasks
tasks = {
"MTurk": fetch_MTurk(),
"MEN": fetch_MEN(),
"WS353": fetch_WS353(),
"RG65":fetch_RG65(),
"RW":fetch_RW(),
"SIMLEX999": fetch_SimLex999(),
"TR9856":fetch_TR9856()
}
result = np.zeros((7,7))
# Print sample data
#for name, data in iteritems(tasks):
# print("Sample data from {}: pair \"{}\" and \"{}\" is assigned score {}".format(name, data.X[0][0], data.X[0][1], data.y[0]))
# Calculate results using helper function
i = 0
def test_MTurk_fetcher():
data = fetch_MTurk()
assert (len(data.y) == len(data.X) == 287)
assert (10.0 >= data.y.max() >= 9)