-
Notifications
You must be signed in to change notification settings - Fork 0
/
semantic_similarity_prototype.py
379 lines (329 loc) · 13.6 KB
/
semantic_similarity_prototype.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
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
import spacy
import re
import numpy
import pandas
import sense2vec
import math
from math import pi, sin, cos
import matplotlib.pyplot as plt
from gensim.models import Word2Vec
from sklearn.cluster import KMeans, Birch, AgglomerativeClustering
from itertools import combinations
nlp = spacy.load('en')
# Download glove word2vec from http://nlp.stanford.edu/projects/glove/
model = Word2Vec.load('glove.6B.300d_word2vec')
path = "PATH_TO_FOLDER"
neurohub_relationships = eval(open(path + 'neurohub_relationships', 'r').readlines()[0])
neurohub_concepts = eval(open(path + 'neurohub_concepts', 'r').readlines()[0])
companies = {'neurohub':[neurohub_concepts,neurohub_relationships]}
def plot_company(companies):
for c in companies:
list_of_concepts = companies[c][0]
list_of_concepts = [l.lower() for l in list_of_concepts]
relationships = companies[c][1]
#create a cluster
clusters, labels = plot_concepts(list_of_concepts, relationships, c)
def plot_all(companies):
for c in companies:
list_of_concepts = companies[c][0]
list_of_concepts = [l.lower() for l in list_of_concepts]
relationships = companies[c][1]
#create a cluster
clusters, labels = plot_concepts(list_of_concepts, relationships, c)
for cl in clusters:
k = math.floor(math.sqrt(len(clusters[cl])))
name = c + "_" +labels[cl][0]
name = name.replace('/', ':')
subcluster, sublabels = plot_concepts(clusters[cl], relationships, name, k)
cluster_combs = [comb for comb in combinations(clusters.keys(), 2)]
for cc in cluster_combs:
cluster_subdict = {x: clusters[x] for x in clusters if x in cc}
keys = list(cluster_subdict.keys())
name = c + "_rel" + ''.join('_' + str(k) for k in keys)
cluster_0 = clusters[keys[0]]; cluster_1 = clusters[keys[1]]
sublabels_0, subclusters_0 = create_clusters(cluster_0, clusters_to_make=math.floor(math.sqrt(len(cluster_0))))
sublabels_0 = build_abstracted_terms_dictionary(subclusters_0)
sublabels_1, subclusters_1 = create_clusters(cluster_1, clusters_to_make=math.floor(math.sqrt(len(cluster_1))))
sublabels_1 = build_abstracted_terms_dictionary(subclusters_1)
icr_z = find_intercluster_relationships_zoom(subclusters_0, subclusters_1, relationships)
plot_relationship_zoom(cluster_subdict, keys, labels, subclusters_0, sublabels_0, subclusters_1, sublabels_1, icr_z, name)
def plot_relationship_zoom(clusters, keys, labels, subclusters_0, sublabels_0, subclusters_1, sublabels_1, ic_relationships, name):
plt.xlim(0.0, 4)
plt.ylim(0.0, 4)
loc_0 = 1
loc_1 = 3
#left cluster
size = len(clusters[keys[0]])
plt.axis('off')
plt.plot(1,1, 'o', markerfacecolor="None", markeredgecolor='k', markersize=70)
plt.annotate(labels[keys[0]][0], xy=(loc_0,0.35), va="bottom", ha='center', size=8)
#right cluster
size = len(clusters[keys[1]])
plt.axis('off')
plt.plot(3,1, 'o', markerfacecolor="None", markeredgecolor='k', markersize=70)
plt.annotate(labels[keys[1]][0], xy=(loc_1,0.35), va="bottom", ha='center', size=8)
draw_zoom_relationships(ic_relationships, subclusters_0, subclusters_1, loc_0, loc_1)
draw_zoom_cluster(subclusters_0, sublabels_0, loc_0)
draw_zoom_cluster(subclusters_1, sublabels_1, loc_1)
plt.savefig('/Users/Garm/Desktop/images/%s.pdf' % name, bbox_inches='tight')
plt.clf()
def draw_zoom_cluster(subcluster, sublabels, location):
num_clusters = len(subcluster)
colors = plt.cm.Spectral(numpy.linspace(0.25, 0.75, num_clusters))
grid_size = 1
location = location-1
r = 0.5
for k, col in zip(set(sublabels),colors):
x, y = calc_xy_circle(k, grid_size, num_clusters)
plt.plot(x+r+location, y+r, 'o', markerfacecolor=col, markeredgecolor='k', markersize=10)
plt.annotate(sublabels[k][0], xy=(x+r+location,y+r), va='center', ha='center', size=3, fontweight='bold')
def draw_zoom_relationships(relationships, subclusters_0, subclusters_1, loc_0, loc_1):
r = 0.5
loc_0 = loc_0-1
loc_1 = loc_1-1
for rel in relationships:
x1, y1 = calc_xy_circle(rel[0][1], 1, len(eval("subclusters_%s" % rel[0][0])))
x2, y2 = calc_xy_circle(rel[1][1], 1, len(eval("subclusters_%s" % rel[1][0])))
x1 = x1 + eval("loc_%s" % rel[0][0])+r
x2 = x2 + eval("loc_%s" % rel[1][0])+r
y1 = y1 + r; y2 = y2 + r
plt.plot([x1, x2], [y1, y2], color='k', linestyle='-', linewidth=1)
# if rel[0][0] != rel[1][0]:
# plt.annotate(''.join(x + ' ' for x in rel[2]), xy=((x1+x2)/2, (y1+y2)/2), size=3)
def find_intercluster_relationships_zoom(clusters_0, clusters_1, relationships):
intracluster_relationships = []
for relationship in relationships:
pos1 = find_term_loc_in_clusters(relationship[0], clusters_0, clusters_1)
pos2 = find_term_loc_in_clusters(relationship[2], clusters_0, clusters_1)
if pos1 != pos2 and pos1 is not None and pos2 is not None:
ls = [pos1, pos2, relationship]
intracluster_relationships.append(ls)
return intracluster_relationships
def find_term_loc_in_clusters(term, clusters_0, clusters_1):
for c in clusters_0:
if term.lower() in clusters_0[c]:
return 0, c
for c in clusters_1:
if term.lower() in clusters_1[c]:
return 1, c
def plot_concepts(list_of_concepts, relationships, name, clusters_to_make=9):
labels, clusters = create_clusters(list_of_concepts, clusters_to_make=clusters_to_make)
abstracted_terms = build_abstracted_terms_dictionary(clusters)
number_of_labels = len(set(abstracted_terms))
rels = count_intracluster_relationship(clusters, relationships)
draw_relationships(clusters, rels, number_of_labels)
# plt.subplots_adjust(left=0.1, right=0.9, top=0.9, bottom=0.1)
draw_clusters(clusters, abstracted_terms, number_of_labels)
plt.axis('off')
plt.savefig('/Users/Garm/Desktop/images/%s.pdf' % name, bbox_inches='tight')
plt.tight_layout()
plt.clf()
return clusters, abstracted_terms
def draw_clus(clusters, abstracted_terms, number_of_labels):
draw_clusters(clusters, abstracted_terms, number_of_labels)
plt.axis('off')
plt.savefig('/Users/Garm/Desktop/%s.pdf' % name, bbox_inches='tight')
plt.clf()
def draw_relationships(clusters, rels, number_of_labels):
grid_size = math.ceil(math.sqrt(number_of_labels))
for rel in rels:
x1, y1 = calc_xy_circle(rel[0], grid_size, number_of_labels)
x2, y2 = calc_xy_circle(rel[1], grid_size, number_of_labels)
width = rels[rel]
plt.plot([x1, x2], [y1, y2], color='k', linestyle='-', linewidth=1+math.sqrt(width))
def draw_clusters(clusters, labels, number_of_labels):
colors = plt.cm.Spectral(numpy.linspace(0.25, 0.75, number_of_labels))
grid_size = math.ceil(math.sqrt(number_of_labels))
plt.xlim(0.05, grid_size+0.05)
plt.ylim(0.05, grid_size+0.05)
for k, col in zip(set(labels),colors):
x, y = calc_xy_circle(k, grid_size, number_of_labels)
plt.plot(x, y, 'o', markerfacecolor=col, markeredgecolor='k', markersize=30+math.sqrt(len(clusters[k]))*13)
plt.annotate(labels[k][0], xy=(x,y), va='center', ha='center', size=14, fontweight='bold')
def calc_xy_circle(i, grid_size, number_of_labels):
if number_of_labels <3: number_of_labels = 3
cx = cy = grid_size/2
r = grid_size/2.8
rad = 2 * pi / number_of_labels * i
x = cx + r * cos(rad)
y = cy + r * sin(rad)
return x, y
def calc_xy(i, grid_size):
x = i%grid_size
y = math.floor(i/grid_size)
return x+(grid_size/10),y+(grid_size/10)
def calc_x_y(i, number_of_labels):
k = number_of_labels + i
number_of_labels = number_of_labels * 2
if i%4 == 0:
x = number_of_labels - k
y = number_of_labels - k
if i%3 == 0:
x = k
y = number_of_labels - k
if i%2 == 0:
x = number_of_labels - k
y = k
else:
x = k
y = k
if x == 0: x +=1
if y == 0: y +=1
return x, y
def create_clusters(list_of_words, clusters_to_make=9):
df, labels_array = build_word_matrices(list_of_words, kind='twod')
linkage = 'ward'
ac = AgglomerativeClustering(linkage=linkage, n_clusters=clusters_to_make)
ac.fit(df)
cluster_labels = ac.labels_
cluster_to_words = find_word_clusters(labels_array, cluster_labels)
print_clusters(cluster_to_words)
return cluster_labels, cluster_to_words
# birch = Birch(n_clusters=None, threshold=1.05)
# birch.fit(df)
# cluster_labels = birch.labels_
# cluster_to_words = find_word_clusters(labels_array, cluster_labels)
# print_clusters(cluster_to_words)
# kmeans_model = KMeans(init='k-means++', n_clusters=clusters_to_make, n_init=10, max_iter=10000)
# kmeans_model.fit(df)
# cluster_labels = kmeans_model.labels_
# cluster_inertia = kmeans_model.inertia_
# cluster_to_words = find_word_clusters(labels_array, cluster_labels)
# print_clusters(cluster_to_words)
def build_abstracted_terms_dictionary(clusters):
abstracted_terms = autovivify_list()
for cls in clusters:
if clusters[cls]:
print(cls)
abstracted_terms[cls].append(find_cluster_abstracted_term(clusters[cls]))
else:
abstracted_terms[cls].append(' ')
return abstracted_terms
def find_cluster_abstracted_term(cluster):
gensim_words = split_gensim_words(cluster)
gensim_words = [w for w in gensim_words if not nlp.is_stop(w)]
gensim_vector = sum(model[w] for w in gensim_words)
if gensim_words is not 0 and gensim_vector is not 0:
most_similar = model.most_similar([gensim_vector])
sim = [word[0] for word in most_similar if word[0] in gensim_words]
if not sim:
return get_closest_word_slow(gensim_vector, gensim_words)[1]
else:
return sim[0]
else:
return ' '
def get_closest_word_slow(vector, words):
cosine_similarity = [0, 'n/a']
for word in words:
new_cos_sim = numpy.dot(vector, model[word])/(numpy.linalg.norm(gensim_vector)* numpy.linalg.norm(model[word]))
if new_cos_sim > cosine_similarity[0]:
cosine_similarity = [new_cos_sim, word]
return cosine_similarity
def split_gensim_words(cluster):
new_words = []
for w in cluster:
try:
if model[w].any():
new_words.append(w)
except:
new_words
words = w.split(' ')
for word in words:
try:
if model[word].any():
new_words.append(word)
except:
new_words
return new_words
def build_word_matrices(list_of_words, kind="twod"):
if kind == "twod":
return build_twod_word_matrices(list_of_words)
elif kind == "vector":
return build_word_vector_matrices(list_of_words)
# requires numpy, pandas, spacy as nlp
def build_twod_word_matrices(list_of_words):
# list_of_words = preprocess_words(list_of_words)
labels = [a for a in list_of_words]
matrix_length = len(list_of_words)
numpy_matrix = numpy.zeros((matrix_length,matrix_length))
pandas_matrix = pandas.DataFrame(numpy_matrix, index=labels, columns=labels)
for row_idx, (row_name, row) in enumerate(pandas_matrix.iterrows()):
for col_idx, col in enumerate(pandas_matrix.columns):
if pandas_matrix.iloc[row_idx][col_idx] == 0:
word1 = nlp(row.name)
word2 = nlp(col)
pandas_matrix.iloc[row_idx][col_idx] = pandas_matrix.iloc[col_idx][row_idx] = word1.similarity(word2)
return pandas_matrix, labels
def build_word_vector_matrices(list_of_words):
# list_of_words = preprocess_words(list_of_words)
numpy_arrays = []
labels_array = []
for word in list_of_words:
labels_array.append(word)
numpy_arrays.append( numpy.array([float(i) for i in nlp.vocab[word].vector]) )
return numpy.array( numpy_arrays ), labels_array
def build_word_vector_matrices_sense2vec(list_of_words):
# list_of_words = preprocess_words(list_of_words)
numpy_arrays = []
labels_array = []
for word in list_of_words:
labels_array.append(word)
numpy_arrays.append( numpy.array([float(i) for i in get_sense2vec_model(word, "NOUN")]) )
return numpy.array( numpy_arrays ), labels_array
#Support stuff
def convert(name):
s1 = re.sub('(.)([A-Z][a-z]+)', r'\1 \2', name)
return re.sub('([a-z0-9])([A-Z])', r'\1 \2', s1).lower()
def preprocess_words(list_of_words):
new_list_of_words = []
for word in list_of_words:
new_list_of_words.append(convert(word))
return new_list_of_words
def find_word_clusters(labels_array, cluster_labels):
cluster_to_words = autovivify_list()
for c, i in enumerate(cluster_labels):
cluster_to_words[ i ].append( labels_array[c] )
return cluster_to_words
def print_clusters(clusters):
for c in clusters:
print(str(c) + ": " + str(clusters[c]))
def get_gensim_model(word):
try:
return model[word]
except:
return numpy.zeros(300, dtype='float32')
def get_sense2vec_model(word, word_type):
try:
return sense2vec_model["%s|%s" % (word, word_type)][1]
except:
return numpy.zeros(128, dtype='float32')
class autovivify_list(dict):
def __missing__(self, key):
value = self[key] = []
return value
def __add__(self, x):
if not self and isinstance(x, Number):
return x
raise ValueError
def __sub__(self, x):
if not self and isinstance(x, Number):
return -1 * x
raise ValueError
def count_intracluster_relationship(clusters, relationships):
ir = find_intracluster_relationships(clusters, relationships)
ir_unique = {tuple(x):ir.count(x) for x in ir}
return ir_unique
def find_intracluster_relationships(clusters, relationships):
intracluster_relationships = []
for relationship in relationships:
pos1 = find_term_loc_in_cluster(relationship[0], clusters)
pos2 = find_term_loc_in_cluster(relationship[2], clusters)
if pos1 != pos2 and pos1 is not None and pos2 is not None:
ls = [pos1, pos2]
ls.sort()
intracluster_relationships.append(ls)
return intracluster_relationships
def find_term_loc_in_cluster(term, clusters):
for c in clusters:
if term.lower() in clusters[c]:
return c