class GetMetricEmbedding:
def __init__(self):
self.graph = Graph()
self.graph.e2idx_file = "../../MyData/DBO/All/entity2id.txt"
self.graph.r2idx_file = "../../MyData/DBO/All/relation2id.txt"
self.graph.triple2idx_file = "../../MyData/DBO/All/triple2id.txt"
self.graph.load_data()
self.graph.load_er_embedding()
def split_cons(self, cons, direction):
if direction == "left":
first_r_name, first_e_name = cons.split()
else:
first_e_name, first_r_name = cons.split()
return first_e_name, first_r_name
def get_metric(self, r_name1, direction1, r_name2, direction2):
example_file_path = "./TwoCons_eval/{}_{}/2consQandA.txt".format(r_name1.replace("dbo:", ""),
r_name2.replace("dbo:", ""))
metric_file = "./TwoCons_eval/{}_{}/metric.txt".format(r_name1.replace("dbo:", ""),
r_name2.replace("dbo:", ""))
first_e_idx_list = []
first_r_idx_list = []
second_e_idx_list = []
second_r_idx_list = []
answer_idx_list = []
with open(example_file_path, 'r', encoding="UTF-8") as f:
all_lines = f.readlines()
cnt = 0
while True:
if cnt >= len(all_lines) or all_lines[cnt].strip() == "":
break
first_e_name, first_r_name = self.split_cons(all_lines[cnt], direction1)
first_e_idx_list.append(self.graph.e_name2id[first_e_name])
first_r_idx_list.append(self.graph.r_name2id[first_r_name])
cnt += 1
second_e_name, second_r_name = self.split_cons(all_lines[cnt], direction2)
second_e_idx_list.append(self.graph.e_name2id[second_e_name])
second_r_idx_list.append(self.graph.r_name2id[second_r_name])
cnt += 1
answer_name = all_lines[cnt].split()
answer_idx_list.append([self.graph.e_name2id[name] for name in answer_name])
cnt += 1
start_time = time.time()
map_r, mrr_r, hits10_r, mean_rank = self.graph.ekg.get_map_mrr_hits10_meanRank_embed_2cons(
first_e_idx_list, first_r_idx_list, direction1,
second_e_idx_list, second_r_idx_list, direction2,
answer_idx_list)
end_time = time.time()
with open(metric_file, 'w', encoding="UTF-8") as f:
f.write("map:{}\n".format(map_r))
f.write("mrr:{}\n".format(mrr_r))
f.write("hits10:{}\n".format(hits10_r))
f.write("mean_rank:{}\n".format(mean_rank))
f.write("average_time:{}\n".format((end_time - start_time) / len(first_e_idx_list)))
t_idx_list.append(e_idx_list[1:])
return h_idx_list, t_idx_list
search_folder = "../../MyData/EmbedDBO/"
graph = Graph()
graph.e2idx_file = search_folder + "entity2id.txt"
graph.r2idx_file = search_folder + "relation2id.txt"
graph.triple2idx_file = search_folder + "triple2id.txt"
if transe_embed is True:
transe_embed = False
graph.load_data()
graph.load_er_embedding()
r_name_list = []
for sparql in eaqs:
sp = SparqlParser(sparql=sparql)
sp.parse_sparql()
for relation in sp.r_name_list:
r_name_list.append(relation)
r_name_list = list(set(r_name_list))
final_map = 0.0
final_hits10 = 0.0
final_mrr = 0.0
final_time = 0