def get_top_k_match(k, source, targets, source_embeddings, target_embeddings):
result_dict_average = {}
result_dict_average_tfidf = {}
result_dict_sum = {}
for t in targets:
distance_average = euclidean_distances(
vector_averaging(source.split(" "), source_embeddings, DIMENSION),
vector_averaging(t.split(" "), target_embeddings, DIMENSION))[0][0]
distance_average_tfidf = euclidean_distances(
vector_averaging_with_tfidf(source.split(" "), source_embeddings,
cs_word2weight, DIMENSION),
vector_averaging_with_tfidf(t.split(" "), target_embeddings,
java_word2weight, DIMENSION))[0][0]
# distance_sum = euclidean_distances(vector_summing(source.split(" "),source_embeddings,DIMENSION),vector_summing(t.split(" "),target_embeddings,DIMENSION))[0][0]
# distance_sum_tfidf = euclidean_distances(vector_summing_with_tfidf(source.split(" "),source_embeddings,cs_word2weight,DIMENSION),vector_averaging_with_tfidf(t.split(" "),target_embeddings,java_word2weight,DIMENSION))[0][0]
result_dict_average[t] = distance_average
result_dict_average_tfidf[t] = distance_average_tfidf
# result_dict_sum[t] = distance_sum
sorted_result_average = sorted(result_dict_average.items(),
key=operator.itemgetter(1))
sorted_result_average_tfidf = sorted(result_dict_average_tfidf.items(),
key=operator.itemgetter(1))
# sorted_result_sum = sorted(result_dict_sum.items(), key=operator.itemgetter(1))
return sorted_result_average[:
k], sorted_result_average_tfidf[:
k] #sorted_result_sum[:k]
def print_phrase_cosine_similarity(phrases,cs_vectors,java_vectors):
for p in phrases:
for p2 in phrases:
cs_vec = vector_averaging(p.split(" "),cs_vectors)
java_vec = vector_averaging(p2.split(" "),java_vectors)
# cs_vec_tfidf = vector_averaging_with_tfidf(p.split(" "),cs_vectors,cs_word2weight)
# java_vec_tfidf = vector_averaging_with_tfidf(p.split(" "),java_vectors,java_word2weight)
cos_sim = cosine_similarity(cs_vec,java_vec)[0][0]
# cos_sim_tfidf = cosine_similarity(cs_vec_tfidf,java_vec_tfidf)[0][0]
with open("cosine_sim_phrase.csv","a") as f:
line = p + "," + p2 + "," + str(cos_sim)
f.write(line + "\n")
def get_top_k_match(k, source, targets, source_embeddings,target_embeddings ):
result_dict_average = {}
result_dict_average_tfidf = {}
result_dict_sum = {}
for t in targets:
distance_average = euclidean_distances(vector_averaging(source.split(" "),source_embeddings,DIMENSION),vector_averaging(t.split(" "),target_embeddings,DIMENSION))[0][0]
distance_average_tfidf = euclidean_distances(vector_averaging_with_tfidf(source.split(" "),source_embeddings,cs_word2weight,DIMENSION),vector_averaging_with_tfidf(t.split(" "),target_embeddings,java_word2weight,DIMENSION))[0][0]
# distance_sum = euclidean_distances(vector_summing(source.split(" "),source_embeddings,DIMENSION),vector_summing(t.split(" "),target_embeddings,DIMENSION))[0][0]
# distance_sum_tfidf = euclidean_distances(vector_summing_with_tfidf(source.split(" "),source_embeddings,cs_word2weight,DIMENSION),vector_averaging_with_tfidf(t.split(" "),target_embeddings,java_word2weight,DIMENSION))[0][0]
result_dict_average[t] = distance_average
result_dict_average_tfidf[t] = distance_average_tfidf
# result_dict_sum[t] = distance_sum
sorted_result_average = sorted(result_dict_average.items(), key=operator.itemgetter(1))
sorted_result_average_tfidf = sorted(result_dict_average_tfidf.items(), key=operator.itemgetter(1))
# sorted_result_sum = sorted(result_dict_sum.items(), key=operator.itemgetter(1))
return sorted_result_average[:k], sorted_result_average_tfidf[:k] #sorted_result_sum[:k]
# top_5_java = java_vectors.similar_by_vector(cs_vectors[cs_k], topn=5)
# # print top_5_java
# relevant_list = list()
# for element in top_5_java:
# if element[0] == cs_k:
# relevant_list.append(1)
# else:
# relevant_list.append(0)
# avg_p = average_precision(relevant_list)
# avg_p_list_cs.append(relevant_list)
# # print avg_p
# precision_list_cs.append(avg_p)
# except Exception as e:
# print e
# print avg_p_list_cs
# print "MAP CS : " + str(mean_average_precision(avg_p_list_cs))
print expr_cs
print expr_java
predict_average = cosine_similarity(vector_averaging(expr_cs.split(" "),cs_vectors,DIMENSION),vector_averaging(expr_java.split(" "),java_vectors,DIMENSION))[0][0]
predict_average_tfidf = cosine_similarity(vector_averaging_with_tfidf(expr_cs.split(" "),cs_vectors,cs_word2weight,DIMENSION),vector_averaging_with_tfidf(expr_java.split(" "),java_vectors,java_word2weight,DIMENSION))[0][0]
new_row = list()
new_row.append(str(predict_average))
new_row.append(str(predict_average_tfidf))
with open("./evaluation_result/keywords_result_11_20_include_functions.csv","a") as f:
f.write(",".join(new_row) + "\n")
# C#
# diff_1 = process_diff_srcml(row[3],0)
# diff_1 = process_source_code(row[3],0)
diff_1 = process_diff_srcml2(row[3])
# Java
# diff_2 = process_diff_srcml(row[5],1)
# diff_2 = process_source_code(row[5],0)
diff_2 = process_diff_srcml2(row[5])
print("diff 1 : " + diff_1)
print("diff 2 : " + diff_2)
# if i > 0 and i != 40:
# diff_cs_list.append(process_source_code(row[5]))
predict_average = cosine_similarity(
vector_averaging(diff_1.split(" "), cs_vectors),
vector_averaging(diff_2.split(" "), java_vectors))[0][0]
predict_average_tfidf = cosine_similarity(
vector_averaging_with_tfidf(diff_1.split(" "), cs_vectors,
cs_word2weight),
vector_averaging_with_tfidf(diff_2.split(" "), java_vectors,
java_word2weight))[0][0]
# predict = doc2vec_model.docvecs.similarity_unseen_docs(doc2vec_model,diff_1.split(" "),diff_2.split(" "))
print(predict_average)
print(predict_average_tfidf)
print(row[7])
new_row = list()
new_row.append(row[0])
new_row.append(row[1])
new_row.append(row[7])
new_row.append(row[8])
# else:
# relevant_list.append(0)
# avg_p = average_precision(relevant_list)
# avg_p_list_cs.append(relevant_list)
# # print avg_p
# precision_list_cs.append(avg_p)
# except Exception as e:
# print e
# print avg_p_list_cs
# print "MAP CS : " + str(mean_average_precision(avg_p_list_cs))
print expr_cs
print expr_java
predict_average = cosine_similarity(
vector_averaging(expr_cs.split(" "), cs_vectors, DIMENSION),
vector_averaging(expr_java.split(" "), java_vectors,
DIMENSION))[0][0]
predict_average_tfidf = cosine_similarity(
vector_averaging_with_tfidf(expr_cs.split(" "), cs_vectors,
cs_word2weight, DIMENSION),
vector_averaging_with_tfidf(expr_java.split(" "), java_vectors,
java_word2weight, DIMENSION))[0][0]
new_row = list()
new_row.append(str(predict_average))
new_row.append(str(predict_average_tfidf))
with open(
"./evaluation_result/keywords_result_11_20_include_functions.csv",
"a") as f:
with codecs.open("./codelabel/codelabel_" + PROJECT + ".csv","r") as f_csv:
reader = csv.reader(f_csv)
for i,row in enumerate(reader):
print("###############################")
# if i == 40:
print("id : " + row[1])
diff_1 = process_source_code(row[3],0)
diff_2 = process_source_code(row[5],0)
print("diff 1 : " + diff_1)
print("diff 2 : " + diff_2)
# if i > 0 and i != 40:
# diff_cs_list.append(process_source_code(row[5]))
predict_average = cosine_similarity(vector_averaging(diff_1.split(" "),cs_vectors,DIMENSION),vector_averaging(diff_2.split(" "),java_vectors,DIMENSION))[0][0]
predict_average_tfidf = cosine_similarity(vector_averaging_with_tfidf(diff_1.split(" "),cs_vectors,cs_word2weight,DIMENSION),vector_averaging_with_tfidf(diff_2.split(" "),java_vectors,java_word2weight,DIMENSION))[0][0]
# predict = doc2vec_model.docvecs.similarity_unseen_docs(doc2vec_model,diff_1.split(" "),diff_2.split(" "))
print(predict_average)
print(predict_average_tfidf)
print(row[7])
new_row = list()
new_row.append(row[0])
new_row.append(row[1])
new_row.append(row[7])
new_row.append(row[8])
new_row.append(row[9])
new_row.append(str(predict_average))
new_row.append(str(predict_average_tfidf))
with open("./codelabel_result/bi2vec_" + PROJECT +"_10.csv","a") as f:
f.write(",".join(new_row) + "\n")