-
Notifications
You must be signed in to change notification settings - Fork 0
/
clustering.py
300 lines (250 loc) · 12.9 KB
/
clustering.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
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
# -*- coding: utf8 -*-
import MySQLdb
import gensim
import yaml
import logging
import logging.config
import codecs
import random
import unicodecsv
import numpy as np
from time import time
from matplotlib import pyplot as plt
from scipy.cluster.hierarchy import dendrogram, linkage, fcluster
from sklearn import metrics
from sklearn.cluster import MiniBatchKMeans, KMeans
from ConfigParser import SafeConfigParser
def generate_embedding_file(embedding_model_name, embedding_model,
embedding_type, items, cluster_label_ground_truth_file=None):
"""Produce a file containing the embeddings of given words.
Produce a file containing the clustering ground truth (labels) of given words."""
embedding_file = 'data/{}/embeddings/{}_{}.npy'.format(dataset, embedding_model_name, embedding_type)
embedding_label_file = 'data/{}/embeddings/{}_{}_labels.txt'.format(dataset, embedding_model_name, embedding_type)
embedding_out = codecs.open(embedding_label_file, 'w', encoding='utf-8')
embeddings = []
if cluster_label_ground_truth_file:
cluster_labels = []
for _ in xrange(len(items[0]) - 1):
cluster_labels.append([])
counter = 0
for item in items:
if item[0] in embedding_model and -1 not in item[1:]:
counter += 1
embeddings.append(embedding_model[item[0]])
embedding_out.write(u'{}\n'.format(item[0]))
for i, x in enumerate(item[1:]):
cluster_labels[i].append(str(x))
np.save(embedding_file, np.array(embeddings))
embedding_out.close()
logger.info('{} {} saved at {}'.format(counter, embedding_type, embedding_file))
logger.info('embedding labels saved at {}'.format(embedding_label_file))
if cluster_label_ground_truth_file:
cluster_label_out = codecs.open(cluster_label_ground_truth_file, 'w', encoding='utf-8')
logger.info('cluster label ground truth saved at {}'.format(cluster_label_ground_truth_file))
for i in xrange(len(cluster_labels)):
cluster_label_out.write(u'{}\n'.format(','.join(cluster_labels[i])))
logger.info('cluster group {}: {} labels'.format(i, len(cluster_labels[i])))
cluster_label_out.close()
def generate_embedding_file_from_csv(embedding_model_name, embedding_type, csv_file):
"""Read entities or relations from a csv file and look up their embeddings given an embedding model."""
items = []
csv_in = open(csv_file)
csv_reader = unicodecsv.reader(csv_in)
for row in csv_reader:
row_len = len(row)
if row_len:
item = [row[1].strip()]
for x in row[2:]:
item.append(int(x))
items.append(item)
logger.info('{} items retrieved'.format(len(items)))
cluster_label_ground_truth_file = None
if len(items[0]) > 1:
cluster_label_ground_truth_file = \
'data/evaluation/clustering/{}_ground_truth.txt'.format(embedding_type)
embedding_model_file = 'data/{}/embeddings/{}'.format(dataset, embedding_model_name)
embedding_model = gensim.models.Word2Vec.load_word2vec_format(embedding_model_file, binary=True)
generate_embedding_file(embedding_model_name, embedding_model, embedding_type,
items, cluster_label_ground_truth_file)
def generate_embedding_file_from_mysql(embedding_model_name, result_n):
"""Read entities and relations from a MySQL db and look up their embeddings given an embedding model."""
def select_sql(table_name='svo'):
return 'SELECT DISTINCT subject_head, predicate_canonical, object_head FROM {} ORDER BY id'.format(table_name)
conn = MySQLdb.connect(**mysql_config)
cur = conn.cursor()
try:
cur.execute(select_sql())
except MySQLdb.Error, e:
try:
logger.error("MySQL Error [{}]: {}".format(e.args[0], e.args[1]))
except IndexError:
logger.error("MySQL Error: {}".format(str(e)))
else:
sql_results = cur.fetchall()
logger.info('{} triples retrieved.'.format(len(sql_results)))
if result_n > len(sql_results):
logger.error('There are not so many results returned from the database!')
exit()
# Collect distinct entities and relations.
entities, relations = set(), set()
# Randomly pick rows.
rows = random.sample(xrange(len(sql_results)), result_n)
for i in rows:
row = sql_results[i]
s = row[0].strip()
p = row[1].strip()
o = row[2].strip()
if s:
entities.add((s.replace(' ', '_'),))
if p:
relations.add((p.replace(' ', '_'),))
if o:
entities.add((o.replace(' ', '_'),))
logger.info('{} entities retrieved'.format(len(entities)))
logger.info('{} relations retrieved'.format(len(relations)))
embedding_model_file = 'data/{}/embeddings/{}'.format(dataset, embedding_model_name)
embedding_model = gensim.models.Word2Vec.load_word2vec_format(embedding_model_file, binary=True)
generate_embedding_file(embedding_model_name, embedding_model, 'entities', list(entities))
generate_embedding_file(embedding_model_name, embedding_model, 'relations', list(relations))
finally:
cur.close()
conn.close()
def agglomerative_clustering(embedding_model_name, embedding_type, cluster_label_ground_truth_file,
cluster_n, method='ward', metric='euclidean', plot=False):
embedding_file = 'data/{}/embeddings/{}_{}.npy'.format(dataset, embedding_model_name, embedding_type)
embeddings = np.load(embedding_file)
logger.info('Loaded embeddings from {}'.format(embedding_file))
# Start clustering.
logger.info('Start clustering ({}, {})...'.format(cluster_n, method))
t0 = time()
clustering = linkage(embeddings, method=method, metric=metric)
logger.info('Clustering time: {}s'.format(time() - t0))
embedding_labels = []
embedding_label_file = 'data/{}/embeddings/{}_{}_labels.txt'.format(dataset, embedding_model_name, embedding_type)
embedding_label_in = codecs.open(embedding_label_file)
for row in embedding_label_in:
if row:
label = row.strip()
if label:
embedding_labels.append(label)
embedding_label_in.close()
cluster_label_prediction = fcluster(clustering, cluster_n, criterion='maxclust') # 1-based index
# logger.info('Cluster label prediction: {}'.format(cluster_label_prediction))
clusters_agg = {}
for i in xrange(len(cluster_label_prediction)):
clusters_agg.setdefault(cluster_label_prediction[i] - 1, []).append(i)
clustering_clusters_file = 'data/{}/clustering/{}_clusters.txt'.format(dataset, embedding_type)
cluster_out = codecs.open(clustering_clusters_file, 'w')
for i in xrange(len(clusters_agg)):
cluster_out.write(u'{}\n'.format(','.join([embedding_labels[j] for j in clusters_agg[i]])))
cluster_out.close()
logger.info('Clustering labels saved at {}'.format(clustering_clusters_file))
if cluster_label_ground_truth_file:
# Read cluster label ground truth
cluster_label_ground_truth = []
with open(cluster_label_ground_truth_file) as f:
for line in f:
if line:
cluster_label_ground_truth.append(map(int, line.strip().split(',')))
# Compute Ajusted Rand Index
for i in xrange(len(cluster_label_ground_truth)):
ari = metrics.adjusted_rand_score(cluster_label_ground_truth[i], cluster_label_prediction)
logger.info('Ajusted Rand Index for cluster group {}: {}'.format(i, ari))
ami = metrics.adjusted_mutual_info_score(cluster_label_ground_truth[i], cluster_label_prediction)
logger.info('Ajusted Mutual Information Score for cluster group {}: {}'.format(i, ami))
chv = metrics.homogeneity_completeness_v_measure(cluster_label_ground_truth[i], cluster_label_prediction)
logger.info('V-measure score for cluster group {}: {}'.format(i, chv))
# Compute Silhouette Coefficient
t0 = time()
sc_score = metrics.silhouette_score(embeddings, cluster_label_prediction, metric=metric)
logger.info('Silhouette Coefficient: {}'.format(sc_score))
logger.info('SC computation time: {}s'.format(time() - t0))
if plot:
plt.rc('lines', linewidth=2)
plt.figure()
plt.title('{} Clustering'.format('Relation'), fontsize=28)
plt.yticks([])
dendrogram(
clustering,
leaf_rotation=90., # rotates the x axis labels
leaf_font_size=14., # font size for the x axis labels
labels=embedding_labels
)
plt.gcf().subplots_adjust(bottom=0.25)
plt.show()
# plt.savefig('data/{}/{}_clustering_dendrogram.png'.format(dataset, type), dpi=300)
return sc_score
def kmeans(embedding_model, embedding_type, cluster_n, cluster_label_ground_truth_file):
embedding_file = 'data/{}/embeddings/{}_{}.npy'.format(dataset, embedding_model, embedding_type)
embeddings = np.load(embedding_file)
logger.info('Loaded embeddings from {}'.format(embedding_file))
# Start clustering.
logger.info('Start clustering ({})...'.format(cluster_n))
k_means = KMeans(init='k-means++', n_clusters=cluster_n, n_init=10)
t0 = time()
k_means.fit(embeddings)
logger.info('Clustering time: {}s'.format(time() - t0))
cluster_label_prediction = k_means.labels_
logger.info('Cluster label prediction: {}'.format(cluster_label_prediction))
# Read cluster label ground truth
cluster_label_ground_truth = []
with open(cluster_label_ground_truth_file) as f:
for line in f:
if line:
cluster_label_ground_truth.append(map(int, line.strip().split(',')))
# Compute Ajusted Rand Index
for i in xrange(len(cluster_label_ground_truth)):
logger.info('Cluster label ground truth: {}'.format(cluster_label_ground_truth[i]))
ari = metrics.adjusted_rand_score(cluster_label_ground_truth[i], cluster_label_prediction)
logger.info('Ajusted Rand Index for cluster group {}: {}'.format(i, ari))
# Compute Silhouette Coefficient
t0 = time()
sc_score = metrics.silhouette_score(embeddings, cluster_label_prediction, metric='euclidean')
logger.info('Silhouette Coefficient: {}'.format(sc_score))
logger.info('SC computation time: {}s'.format(time() - t0))
return sc_score
def cross_validate(embedding_type, methods, cluster_numbers, result_file):
with open(result_file, 'w') as r_out:
for method in methods:
for cluster_n in cluster_numbers:
r_out.write('{}, {}, {}\n'.format(
method, cluster_n, agglomerative_clustering(embedding_type, cluster_n, method)))
if __name__ == '__main__':
with open('config/logging_config.yaml') as f:
logging.config.dictConfig(yaml.load(f))
logger = logging.getLogger('clustering')
dataset = 'pmc_c-h'
db = 'bio-kb'
# Read MySQL configurations.
parser = SafeConfigParser()
parser.read('config/mysql_config.ini')
mysql_config = {
'host': parser.get('MySQL', 'host'),
'user': parser.get('MySQL', 'user'),
'passwd': parser.get('MySQL', 'passwd'),
'db': db
}
embedding_model_name = 'scikb_directed'
# embedding_model_name = 'word2vec'
# embedding_model_name = 'pmc_w2v'
# embedding_type = 'entities'
embedding_type = 'relations'
# benchmark_file = 'data/evaluation/clustering/{}.csv'.format(embedding_type)
# generate_embedding_file_from_csv(embedding_model_name, embedding_type, benchmark_file)
# generate_embedding_file_from_mysql(embedding_model_name, 100)
# cluster_label_ground_truth_file = 'data/evaluation/clustering/{}_ground_truth.txt'.format(embedding_type)
cluster_label_ground_truth_file = None
# cluster_n: all-3,40; e1-15; e2-7; e3-18
agglomerative_clustering(embedding_model_name, embedding_type, cluster_label_ground_truth_file,
20, method='ward', metric='euclidean', plot=True)
# kmeans(embedding_model_name, 'entities', 26, cluster_label_ground_truth_file)
# Cross validation for clustering parameters.
# clustering_methods = ['single', 'complete', 'average', 'weighted', 'centroid', 'median', 'ward']
# entity_clustering_result_file = 'data/{}/clustering/entity_clustering_results.txt'.format(dataset)
# entity_clustering_cluster_numbers = [2, 5, 10, 20, 50, 100]
# cross_validate('entity_embedding', clustering_methods,
# entity_clustering_cluster_numbers, entity_clustering_result_file)
# relation_clustering_result_file = 'data/{}/clustering/relation_clustering_results.txt'.format(dataset)
# relation_clustering_cluter_numbers = [3000]
# cross_validate('relation_embedding', clustering_methods,
# relation_clustering_cluter_numbers, relation_clustering_result_file)