def generate_statistics_with_weighted_majorityvote_ensamble(
startingpoint=5, endpoint=50, skips=5, iterations=5, topn=10):
results = []
for i in range(startingpoint, endpoint, skips):
for j in range(0, iterations):
if (os.path.isfile(
"weighted_majority_vote_Ensamble_test_results.csv")):
f = open("weighted_majority_vote_Ensamble_test_results.csv",
"a")
else:
f = open("weighted_majority_vote_Ensamble_test_results.csv",
"w")
ensamble_model = BS.boot_strap_aggregator()
dir_path = "questions-words.txt"
right, wrong = ensamble_model.accuracy(dir_path,
number_of_models=i,
predictor_method=3)
res = [[i, topn, right, wrong]]
print(res)
results.append(res)
np.savetxt(f, res, delimiter=',')
f.close()
print('iteration finished')
print(results)
return results
def generate_statistics_weight_based_on_total_oov_ignore_oov_human_similarity(
startingpoint=5, endpoint=50, skips=5, iterations=5, topn=10):
results = []
for i in range(startingpoint, endpoint, skips):
for j in range(0, iterations):
if (os.path.isfile(
"weight_based_on_total_oov_ignore_oov_human_similarity_stats.csv"
)):
f = open(
"weight_based_on_total_oov_ignore_oov_human_similarity_stats.csv",
"a")
else:
f = open(
"weight_based_on_total_oov_ignore_oov_human_similarity_stats.csv",
"w")
ensamble_model = BS.boot_strap_aggregator()
dir_path = "wordsim353.tsv"
spearman_result, pearson_result = ensamble_model.evaluate_word_pairs(
dir_path, number_of_models=i, similarity_model_type=3)
res = [[
i, topn, spearman_result[0], spearman_result[1],
pearson_result[0], pearson_result[1]
]]
print(res)
results.append(res)
np.savetxt(f, res, delimiter=',')
f.close()
print('iteration finished')
print(results)
return results
def oov_test(startingpoint, endpoint, skips, iterations):
results = []
for i in range(startingpoint, endpoint, skips):
for j in range(0, iterations):
if (os.path.isfile("oov_test.csv")):
f = open("oov_test.csv", "a")
else:
f = open("oov_test.csv", "w")
ensamble_model = BS.boot_strap_aggregator()
dir_path = "questions-words.txt"
oov = ensamble_model.oov_test(questions=dir_path,
number_of_models=i)
res = [[i, oov]]
print(res)
results.append(res)
np.savetxt(f, res, delimiter=',')
f.close()
print('iteration finished')
def word_sim_test():
ensamble = EP.simple_ensamble(
generate_array_of_all_trained_model_specification)
dir_path = os.path.dirname(
os.path.realpath(__file__)) + "/TestingSet/wordsim353.tsv"
ensamble.evaluate_word_pairs(dir_path, similarity_model_type=0)
def question_word_test():
ensamble = EP.boot_strap_aggregator(
generate_array_of_all_trained_model_specification())
dir_path = "questions-words.txt"
#ensamble.set_weights([1, 0.5])
ensamble.accuracy(dir_path, predictor_method=3, number_of_models=2)