-
Notifications
You must be signed in to change notification settings - Fork 6
/
Copy pathexport_d2s.py
173 lines (152 loc) · 7.64 KB
/
export_d2s.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
from gensim import models
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.feature_extraction.text import CountVectorizer
from sklearn import metrics
import numpy as np
from gensim.models import word2vec,LdaModel
from gensim.corpora import MmCorpus
import cPickle,os,re,codecs,sys,glob
from categoricalClustDist import categorical_clust_dist,token_to_mat
#from spelling_corrector import correct
def extract_features(in_file='story_clusters.txt',features_file='feature_words.txt', filtered_category_list=[]):
f = codecs.open(in_file,'r',encoding='utf-8',errors='ignore')
clusters = f.read().strip('%%%').split('\n%%%')
docs = []
labels = []
word_length = 0
document_count = 0
for i,cluster in enumerate(clusters):
if (not filtered_category_list) or (cluster.strip().split('\n')[0] in filtered_category_list):
for ln in cluster.strip().split('\n')[1:]:
docs.append(ln)
labels.append(i)
document_count = document_count + 1
word_length = word_length + len(ln.split(' '));
print "raw categories: %d" % len(clusters)
print "document count: %d" % document_count
print "average words: %d" % (word_length / document_count)
vectorizer = CountVectorizer(lowercase=True, stop_words='english', token_pattern=r"(?u)\b[a-zA-Z_][a-zA-Z_]+\b")
vectorizer.fit_transform(docs)
features = vectorizer.get_feature_names()
fw = open(features_file,'w')
fw.write(u'\n'.join(features).encode('utf-8'))
fw.close()
def build_word2vec_index(feature_file='feature_words.txt', word_vecs_file='GoogleNews-vectors-negative300.bin', out_dic='word_vecs.pkl'):
words = open(feature_file).read().strip().split('\n')
dic = dict()
model = word2vec.Word2Vec.load_word2vec_format(word_vecs_file, binary=True, norm_only=False)
for w in words:
try:
v = model[w]
dic[w] = v
except:
continue # whatever
try:
v = model[correct(w)]
dic[w] = v
except:
print w + ' not existed in word2vec model'
continue
fw = open(out_dic,'wb')
cPickle.dump(dic, fw)
fw.close()
def convert_d2_format(in_file='story_clusters.txt', embedding_dic='word_vecs.pkl',embedding_dim_size=300, weighting_type='tf', d2_vocab='story_cluster.d2s.vocab0', d2s_file='story_cluster.d2s',filtered_category_list=[]):
word2vec_dic = cPickle.load(open(embedding_dic,'rb'))
vocab = word2vec_dic.keys()
#clusters = open(in_file).read().strip('%%%').split('\n%%%')
f = codecs.open(in_file,'r',encoding='utf-8',errors='ignore')
clusters = f.read().strip('%%%').split('\n%%%')
docs = []
labels = []
for i,cluster in enumerate(clusters):
if (not filtered_category_list) or (cluster.strip().split('\n')[0] in filtered_category_list):
for ln in cluster.strip().split('\n')[1:]:
docs.append(ln)
labels.append(i)
if weighting_type == 'tf':
vectorizer = CountVectorizer(lowercase=True, stop_words='english', vocabulary=vocab)
if weighting_type == 'tfidf':
vectorizer = TfidfVectorizer(lowercase=True, stop_words='english', vocabulary=vocab)
X = vectorizer.fit_transform(docs)
print X.shape
fw = open(d2s_file,'w')
for i in range(X.shape[0]):
fw.write(str(embedding_dim_size)+'\n')
nonzero_ids = X[i].nonzero()
if (len(nonzero_ids[0])>0):
fw.write(str(len(nonzero_ids[0]))+'\n')
fw.write(' '.join(str(X[i][0,j]) for j in nonzero_ids[1])+'\n')
fw.write(' '.join(str(j+1) for j in nonzero_ids[1])+'\n')
else:
print >> sys.stderr, "empty document found!"
fw.write('1\n')
fw.write('1\n')
fw.write('0\n')
fw.close()
fw = open(d2_vocab,'w')
fw.write(str(embedding_dim_size)+' '+str(len(vocab))+'\n')
fw.write('\n'.join(' '.join(str(v) for v in word2vec_dic[w]) for w in vocab))
fw.close()
def d2clustering_metrics(in_file='story_clusters.txt',in_label_file=None,filtered_category_list=[]):
f = codecs.open(in_file, 'r', encoding='utf-8', errors='ignore')
clusters = f.read().strip('%%%').split('\n%%%')
labels = []
for i,cluster in enumerate(clusters):
if (not filtered_category_list) or (cluster.strip().split('\n')[0] in filtered_category_list):
for ln in cluster.strip().split('\n')[1:]:
labels.append(i)
lines = open(in_label_file).read().strip().split()
d2_labels = [int(i) for i in lines]
print 'number of clusters: %d' % (max(d2_labels)+1)
print np.array(np.sum(token_to_mat(labels), axis=0), dtype='int32')
print np.array(np.sum(token_to_mat(d2_labels), axis=0), dtype='int32')
print 'Homogeneity: %0.3f' % metrics.homogeneity_score(labels, d2_labels)
print 'Completeness: %0.3f' % metrics.completeness_score(labels, d2_labels)
print 'V-measure: %0.3f' % metrics.v_measure_score(labels, d2_labels)
print 'Normalized Mutual Information: %0.3f' % metrics.normalized_mutual_info_score(labels, d2_labels)
print 'Adjusted Mutual Information: %0.3f' % metrics.adjusted_mutual_info_score(labels, d2_labels)
print 'Adjusted Rand Index: %0.3f' % metrics.adjusted_rand_score(labels, d2_labels)
#print 'Categorical Cluster Distance: %0.3f' % categorical_clust_dist(token_to_mat(labels), token_to_mat(d2_labels), method='instance_count')['dist']
if __name__ == '__main__':
dataset = 'story'
#dataset = 'bbc'
#dataset = 'bbc_title'
#dataset = 'bbc_abstract'
#dataset = 'reuters'
#dataset = 'reuters_title'
#dataset = 'bbcsport'
#dataset = 'bbcsport_title'
#dataset = '20newsclean'
#dataset = 'ohsumed_sz25'
vec_dim = 400
word_vecs='word2vec_400_10_10.bin'
#word_vecs='glove_6B_300d.bin'
#word_vecs='GoogleNews-vectors-negative300.bin'
#word_vecs='ohsumed-full_50_20_2.8.bin'
cluster_file = 'acl2017dataset/' + dataset + '_clusters.txt'
vec_dic = dataset + '_word_vecs.pkl'
reuters_r10_categories = ['acq', 'crude', 'earn', 'coffee', 'interest', 'money-fx', 'money-supply', 'ship', 'trade', 'sugar']
if dataset == 'reuters':
category_list = reuters_r10_categories
else:
category_list = []
is_result_avail = False;
for label_file in glob.glob("./"+dataset+"_*_o"):
print '---------------------------------------------------'
print 'Method: D2 Clustering'
print 'Vocabulary Embedding: ' + word_vecs
d2clustering_metrics(in_file=cluster_file, in_label_file=label_file, filtered_category_list=category_list)
is_result_avail = True;
if is_result_avail:
sys.exit(0)
if dataset.startswith('reuters'):
extract_features(in_file=cluster_file, features_file='reuters.terms', filtered_category_list = reuters_r10_categories)
build_word2vec_index(feature_file='reuters.terms', word_vecs_file=word_vecs, out_dic=vec_dic)
convert_d2_format(in_file=cluster_file, embedding_dic=vec_dic, embedding_dim_size=vec_dim, weighting_type='tfidf', d2_vocab=dataset + '_cluster.d2s.vocab0', d2s_file=dataset + '_cluster.d2s', filtered_category_list = reuters_r10_categories)
else:
extract_features(in_file=cluster_file, features_file=dataset + '.terms')
build_word2vec_index(feature_file=dataset + '.terms', word_vecs_file=word_vecs, out_dic=vec_dic)
convert_d2_format(in_file=cluster_file, embedding_dic=vec_dic, embedding_dim_size=vec_dim, weighting_type='tfidf', d2_vocab=dataset + '_cluster.d2s.vocab0', d2s_file=dataset + '_cluster.d2s')
#
#build_word2vec_index()
#convert_d2_format()