/
evaluate.py
207 lines (182 loc) · 8.08 KB
/
evaluate.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
# -*- coding: utf8 -*-
import logging
import logging.config
import yaml
import codecs
import unicodecsv
import scipy.stats
import numpy as np
from cycler import cycler
from matplotlib import pyplot as plt
from gensim.models import Word2Vec
from numpy import arange
from extract_relations import RelationExtractor
def evaluate_extraction(input_file, output_file):
logger = logging.getLogger()
parser_server = 'http://localhost:8084'
count = 0
f_in = codecs.open(input_file, encoding='utf-8')
f_out = open(output_file, 'w')
writer = unicodecsv.writer(f_out)
for line in f_in:
line = line.strip()
if line:
logger.debug(line)
try:
extractor = RelationExtractor(line, parser_server, logger, entity_linking_flag=False)
except:
logger.error(u'Failed to parse the sentence', exc_info=True)
else:
count += 1
extractor.extract_spo()
for relation in extractor.relations:
logger.debug(relation.lemma)
row = [''] * 5
row[0], row[1] = count, line
row[2:5] = relation.lemma
writer.writerow(row)
f_in.close()
f_out.close()
# def evaluate_similarity(method_name, embedding_file):
# logger = logging.getLogger()
#
# logger.info('Loading embeddings from {}...'.format(embedding_file))
# embedding_model = Word2Vec.load_word2vec_format(embedding_file, binary=True)
#
# benchmark_file = 'data/evaluation/similarity/umnsrs_similarity_modified.csv'
# result_file = 'data/evaluation/similarity/{}_similarity_pairs.csv'.format(method_name)
# benchmark_scores = []
# method_scores = []
# f_in = open(benchmark_file)
# f_out = open(result_file, 'w')
# reader = unicodecsv.reader(f_in)
# writer = unicodecsv.writer(f_out)
# next(reader, None)
# for row in reader:
# score, _, term_1, term_2 = row[:4]
# term_1, term_2 = term_1.lower().replace(' ', '_'), term_2.lower().replace(' ', '_')
# score = float(score)
# if term_1 in embedding_model and term_2 in embedding_model:
# sim_score = float(embedding_model.similarity(term_1, term_2))
# logger.debug('{}, {}: {}'.format(term_1, term_2, sim_score))
# writer.writerow([sim_score, term_1, term_2])
# benchmark_scores.append(score)
# method_scores.append(sim_score)
# f_in.close()
# f_out.close()
#
# logger.info('Computing Spearman\'s Rank Correlation...')
# rho, p = scipy.stats.spearmanr(benchmark_scores, method_scores)
# logger.debug('correlation coefficient: {}, p-value: {}'.format(rho, p))
#
# x = np.arange(len(benchmark_scores))
# y1 = np.array(method_scores)
# y1 = (y1 - y1.mean()) / y1.std()
# y2 = np.array(benchmark_scores)
# y2 = (y2 - y2.mean()) / y2.std()
#
# plt.rc('lines', linewidth=2)
# plt.rc('axes', prop_cycle=(cycler('color', ['#E87F4D', '#9CB2B3'])))
# plt.rc('font', **{'size': 24})
# plt.subplot(121)
# plt.plot(x, y1, x, y2)
# plt.xlim([0, np.max(x)])
# plt.xticks([])
# plt.legend([method_name, 'benchmark'])
# plt.subplot(122)
# plt.scatter(y1, y2, s=40, c='#E87F4D', edgecolors='face')
# plt.xlabel(method_name, fontsize=32)
# plt.ylabel('benchmark', fontsize=32)
# plt.xticks([])
# plt.yticks([])
# plt.show()
def generate_entity_pairs(threshold):
input_file = 'data/evaluation/similarity/umnsrs_similarity_modified.csv'
output_file = 'data/evaluation/similarity/entity_pairs.txt'
f_in = open(input_file)
f_out = open(output_file, 'w')
reader = unicodecsv.reader(f_in)
next(reader, None)
for row in reader:
score, _, term_1, term_2 = row[:4]
term_1, term_2 = term_1.lower().split(), term_2.lower().split()
score = float(score)
if score >= threshold and len(term_1) == 1 and len(term_2) == 1:
f_out.write('{},{}\n'.format(term_1[0], term_2[0]))
f_in.close()
f_out.close()
def compute_pair_similarity(benchmark_file, embedding_file, binary_embedding=True):
logger = logging.getLogger()
logger.info('Loading embeddings from {}...'.format(embedding_file))
embedding_model = Word2Vec.load_word2vec_format(embedding_file, binary=binary_embedding)
pair_similarities = {}
with open(benchmark_file) as bf:
for line in bf:
line = line.strip()
if line:
pair = tuple(line.split(','))
term_1, term_2 = pair
if term_1 in embedding_model and term_2 in embedding_model:
sim_score = float(embedding_model.similarity(term_1, term_2))
pair_similarities[pair] = sim_score
accuracy = []
for threshold in arange(0.0, 1.1, 0.1):
similar_pair_count = 0
for pair in pair_similarities:
if pair_similarities[pair] >= threshold:
similar_pair_count += 1
accuracy.append(float(similar_pair_count) / len(pair_similarities))
logger.info('Accuracy: {}'.format(accuracy))
return accuracy
def plot_pair_similarity_results(methods, results, output_figure):
fig, ax = plt.subplots()
line_colors = ['#e8814c', '#9cb2b3', '#e5cb80']
x = arange(0.0, 1.1, 0.1)
for i, y in enumerate(results):
plt.plot(x, y, color=line_colors[i], linestyle='-', linewidth=2, label=methods[i])
tick_font = {'fontname': 'Arial', 'size': '14'}
label_font = {'fontname': 'Arial', 'size': '18'}
ax.set_xlabel('Similarity Score Threshold', **label_font)
ax.set_xlim(0, 1.0)
ax.set_xticks(arange(0.0, 1.1, 0.1))
ax.set_xticklabels(arange(0.0, 1.1, 0.1), **tick_font)
ax.set_ylabel('Accuracy', **label_font)
ax.set_ylim(0, 1.1)
ax.set_yticks(arange(0.0, 1.1, 0.2))
ax.set_yticklabels(arange(0.0, 1.1, 0.2), **tick_font)
ax.legend(methods, fontsize=20, loc='lower left')
ax.grid(alpha=0.6)
fig.savefig('/Users/HanWang/Dropbox/Research/Dissertation/Thesis/figures/{}'.format(output_figure), dpi=300)
if __name__ == '__main__':
with open('config/logging_config.yaml') as f:
logging.config.dictConfig(yaml.load(f))
# Evaluate extraction.
# extraction_input_file = 'data/evaluation/extraction/sentences.txt'
# extraction_outupt_file = 'data/evaluation/extraction/output.csv'
# evaluate_extraction(extraction_input_file, extraction_outupt_file)
# Evaluate pair similarities.
methods = ['SciKB', 'Word2Vec', 'DepW2V']
# Entities
generate_entity_pairs(threshold=1000)
entity_pair_similarity_results = []
entity_pair_benchmark_file = 'data/evaluation/similarity/entity_pairs.txt'
entity_pair_similarity_results.append(
compute_pair_similarity(entity_pair_benchmark_file,
'data/pmc_c-h/embeddings/word2vec'))
# 'data/pmc/embeddings/scikb_directed_58m_triples_min_edge_cnt_two'))
entity_pair_similarity_results.append(
compute_pair_similarity(entity_pair_benchmark_file, 'data/pmc/embeddings/word2vec_58m_triples'))
entity_pair_similarity_results.append(
compute_pair_similarity(entity_pair_benchmark_file, 'data/pmc/embeddings/dep_w2v', False))
plot_pair_similarity_results(methods, entity_pair_similarity_results, 'ch4_entity_pair_similarity_results.eps')
# Relations
relation_pair_similarity_results = []
relation_pair_benchmark_file = 'data/evaluation/similarity/relation_pairs.txt'
relation_pair_similarity_results.append(
compute_pair_similarity(relation_pair_benchmark_file,
'data/pmc/embeddings/scikb_directed_58m_triples_min_edge_cnt_two'))
relation_pair_similarity_results.append(
compute_pair_similarity(relation_pair_benchmark_file, 'data/pmc/embeddings/word2vec_58m_triples'))
relation_pair_similarity_results.append(
compute_pair_similarity(relation_pair_benchmark_file, 'data/pmc/embeddings/dep_w2v', False))
plot_pair_similarity_results(methods, relation_pair_similarity_results, 'ch4_relation_pair_similarity_results.eps')