-
Notifications
You must be signed in to change notification settings - Fork 0
/
sif_embedding.py
452 lines (382 loc) · 20.2 KB
/
sif_embedding.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
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
from sklearn.ensemble._gradient_boosting import np_bool
from src import data_io,params,SIF_embedding
import pickle
import textcleaner as tc
import spacy
import math
import numpy as np
import torch
from sklearn.metrics import davies_bouldin_score,silhouette_score,calinski_harabaz_score
import matplotlib.pyplot as plt
from gensim.sklearn_api import tfidf
from scipy.cluster._hierarchy import cluster_in
from torch import *
import torch.nn as nn
import torch.optim as optim
from pygments.lexer import words
from setuptools.command.dist_info import dist_info
from spacy.syntax.nn_parser import precompute_hiddens
from torch.autograd import Variable
import pandas as pd
import random
from torch.utils.data import DataLoader
import itertools
import nltk
from nltk.corpus import stopwords
import random
from collections import Counter
import string
import re
from CNN_LSTM_model import CNN_LSTM
from CNN_encoder import CNN_Encoder
from Early_Stopping import EarlyStopping
import itertools
from operator import itemgetter
import spacy
import pandas as pd
from gensim.corpora import Dictionary
import textcleaner as tc
from docutils.nodes import section, sidebar
from gensim.models import Word2Vec
from nltk.tokenize import sent_tokenize,word_tokenize
import string
import torch.nn.functional as F
import pickle
from spacy.language import Language
import EncoderRNN as en
import DecoderRNN as dc
from torch import *
from Clustering_experiment import *
from spacy.vocab import Vocab
from spacy.lang.en import English
import spacy
import os
from jqmcvi import base
import pandas as pd
import gensim
from gensim.corpora import Dictionary
from gensim.models import TfidfModel
from gensim.models import Word2Vec
from gensim.models import WordEmbeddingSimilarityIndex
from gensim.similarities import SparseTermSimilarityMatrix, SoftCosineSimilarity, MatrixSimilarity
from gensim.similarities import Similarity
from sklearn.cluster import AffinityPropagation, AgglomerativeClustering, Birch, DBSCAN, SpectralClustering
import pandas as pd
import numpy as np
import nltk
from nltk.tokenize import word_tokenize
from sklearn.metrics.pairwise import cosine_similarity
import os
nlp = spacy.load('en_core_web_sm')
# input Hyper parameter
Word2Vec_path = 'models/Word2vec'
SIF_embedding_path= 'models/SFI'
Cluster_ouput_path = 'Output'
word2vecfile = 'model.txt' # word vector file, can be downloaded from GloVe website
weightfile = 'freqency_dict.txt' # each line is a word and its frequency
sif_file='SIF_embeddings'
cluster_output_file = 'cluster_{}_{}.csv'
cluster_info_file = 'cluster_info.csv'
pure_cluster_file = 'pure_cluster_{}.csv'
noisy_cluster_file = 'noisy_cluster_{}.csv'
weightpara = 1e-3 # the parameter in the SIF weighting scheme, usually in the range [3e-5, 3e-3]
rmpc = 1 # number of principal components to remove in SIF weighting scheme
Desired_cluster = 25
Min_cluster = 2
Binary_threshold = 0.80
Noise_threshold = 0.4
Fixed_K = 15
Power_constant = 0.5
Window_size_for_k = 5
Pure_cluster_threshold = 0.8
Cluster_size = 3
algo='hierarchical'#['k_means','hierarchical']
cluster_scoring_function = 'davies_bouldin_score'#['silhouette_score','davies_bouldin_score','calinski_harabaz_score','dunn']
SIF_embedding_generate = False
pretrained_word2vec = True
test = True
k_calculation = False
def new_lemmatization(sentences:[str],allowed_postags=['NN', 'NNS', 'NNP', 'NNPS', 'VB', 'VBD', 'VBG', 'VBN','VBP', 'VBZ', 'JJ', 'JJR', 'JJS']):#'RB', 'RBR', 'RBS',
texts_out = []
for sent in sentences:
tokens = word_tokenize(sent)
tagged = nltk.pos_tag(tokens)
pos_cleaned_sent = " ".join([token for (token, pos) in tagged if pos in allowed_postags])
doc = nlp(pos_cleaned_sent)
# Extract the lemma for each token and join
texts_out.append(" ".join([token.lemma_ if token.lemma_ not in ['-PRON-'] else '' for token in doc]))
return texts_out
def remove_null_sentence(sentences:[str]):
return [x for x in sentences if x is not '']
def word_tokenizer(sentences:[str]):
sentence=[]
for raw_sentence in sentences:
# If a sentence is empty, skip it
if len(raw_sentence) > 0:
# Otherwise, get a list of words
sentence.append(word_tokenize(raw_sentence))
return sentence
# Return the list of sentences
# so this returns a list of lists
def text_preprocessing(sentences:[str]):
input_text = list(tc.document(sentences).remove_numbers().remove_stpwrds().remove_symbols().lower_all())
lema = new_lemmatization(sentences=input_text)
return lema
def load_dict_from_file(path):
frequency_dict = {}
with open(os.path.join(path,weightfile), 'r') as raw_data:
for item in raw_data:
key,value = item.rstrip().split(' ',1)
frequency_dict[key] = int(value)
return frequency_dict
def save_dict_to_file(word2vec_model,path):
with open(os.path.join(path,weightfile), 'w') as f:
f.writelines('{} {}\n'.format(word, freq.count) for word, freq in word2vec_model.wv.vocab.items())
print('== Dictionary Saved at {} ==='.format(os.path.join(path,weightfile)))
return os.path.join(path,weightfile)
def train_word2vec(Input_sentence, Embedding_size,path):
model = Word2Vec(Input_sentence, size=Embedding_size, min_count=1, workers=8,compute_loss=True) #replace with bigram_sent,sentence for bigram model, unigram evaluation
print(model)
model.train(Input_sentence,total_examples=len(Input_sentence),epochs=10)
#model.init_sims(replace=True)
model.save(os.path.join(path,'Word2Vec.model'))
model.wv.save_word2vec_format(os.path.join(path,'W2V_model.txt'),binary=False)
print("model is trained and saved at {}".format(os.path.join(path,'W2V_model.txt')))
return model
def generate_SFI_embedding(sentences,words,We,weight4ind,params,path,phase='training'):
# load sentences
x, m, = data_io.sentences2idx(sentences, words) # x is the array of word indices, m is the binary mask indicating whether there is a word in that location
w = data_io.seq2weight(x, m, weight4ind) # get word weights
# get SIF embedding
total_embedding = SIF_embedding.SIF_embedding(We, x, w, params,phase=phase) # embedding[i,:] is the embedding for sentence i
np.save(os.path.join(path,sif_file),total_embedding)
print("==========embeddings saved at {} ========".format(os.path.join(path,sif_file)))
return total_embedding
def calculate_cosine_similarity(a,b,df,k=30):
#print("Actual ticket : ",df[index])
similarity_matrix = cosine_similarity(a.reshape(1,-1), b)
index = similarity_matrix[0].argsort()[-k:][::-1]
for i in index:
print(i, "------->", df.iloc[i].values,"---->",similarity_matrix[0][i])
return index
def new_lemmatization(sentences:[str],allowed_postags=['NN', 'NNS', 'NNP', 'NNPS', 'RB', 'RBR', 'RBS', 'VB', 'VBD', 'VBG', 'VBN','VBP', 'VBZ', 'JJ', 'JJR', 'JJS']):
texts_out = []
for sent in sentences:
tokens = word_tokenize(sent)
tagged = nltk.pos_tag(tokens)
pos_cleaned_sent = " ".join([token for (token, pos) in tagged if pos in allowed_postags])
doc = nlp(pos_cleaned_sent)
# Extract the lemma for each token and join
texts_out.append(" ".join([token.lemma_ if token.lemma_ not in ['-PRON-'] else '' for token in doc]))
return texts_out
def remove_null_sentence(sentences:[str]):
return [x for x in sentences if x is not '']
def word_tokenizer(sentences:[str]):
sentence=[]
for raw_sentence in sentences:
# If a sentence is empty, skip it
if len(raw_sentence) > 0:
# Otherwise, get a list of words
sentence.append(word_tokenize(raw_sentence))
return sentence
# Return the list of sentences
# so this returns a list of lists
def text_preprocessing(sentences:[str]):
input_text = list(tc.document(sentences).remove_numbers().remove_stpwrds().remove_symbols().lower_all())
lema = new_lemmatization(sentences=input_text)
return lema
def testing_embedding(total_embedding,words,weight4ind,We,df):
temp=[]
x=input("Enter ticket index : ")
temp.append(x)
preprocessed_sentence = remove_null_sentence(text_preprocessing(sentences=temp))
# load sentences
x, m, = data_io.sentences2idx(preprocessed_sentence,words) # x is the array of word indices, m is the binary mask indicating whether there is a word in that location
w = data_io.seq2weight(x, m, weight4ind) # get word weights
# get SIF embedding
input_embedding = SIF_embedding.SIF_embedding(We, x, w, params,phase='testing') # embedding[i,:] is the embedding for sentence i
index = calculate_cosine_similarity(input_embedding,total_embedding,df)
def clustering_engine(data:np.array,df,file_name,algo:['k_means','hierarchical']='k_means',k=11):
#affinity in Agglomeratigve could be "l1","l2","manhattan", "cosine", or 'precomputed'.
data_frames = []
cluster_ind=[]
if algo == 'k_means':
clusterer = KMeansClusterer(num_means=k, distance=nltk.cluster.util.cosine_distance, avoid_empty_clusters=True,repeats=1,normalise=True)
assigned_clusters = np.array(clusterer.cluster(data, assign_clusters=True))
else:
clusterer = AgglomerativeClustering(n_clusters=k, affinity="cosine", linkage='complete').fit(data)
assigned_clusters = np.array(clusterer.labels_)
for i in range(k):
cluster_index = np.where(assigned_clusters==i)[0]
data_frames.append(df.iloc[cluster_index])
#data_frames.append(pd.DataFrame(list(itertools.chain.from_iterable([df.iloc[j].values[0] for j in cluster_index]))))
cluster_ind.append(cluster_index)
#create a CSV file
clust_info = pd.concat(data_frames, keys=list(range(k)),names=['Cluster_no', 'Row ID']).to_csv(os.path.join(Cluster_ouput_path,file_name.format(k,algo)))
return clusterer,cluster_ind,assigned_clusters
def cluster_overlapping_score(sentence_embeddings,cluster_id,churn_threshold = 0.5):
assigned_cluster_id=cluster_id
total_mean_info= []
for one_cluster_index in assigned_cluster_id:
cluster_mean=[]
a = np.array([sentence_embeddings[j] for j in one_cluster_index])
print(a.shape)
for each_cluster in assigned_cluster_id:
b = np.array([sentence_embeddings[j] for j in each_cluster])
print(cosine_similarity(a,b).shape)
cluster_mean.append(cosine_similarity(a,b).mean())
total_mean_info.append(cluster_mean)
def inter_noise_handling(sentence_embeddings,cluster_id,binary_threshold=.2,noise_threshold=0.5):
efficiency=[]
for i in cluster_id:
one_cluster_index=np.array([sentence_embeddings[j] for j in i])
similarity_matrix=cosine_similarity(one_cluster_index,one_cluster_index)
binay_matrix = np.where(similarity_matrix < binary_threshold, 0, 1)# np.where(similarity_matrix<(similarity_matrix.mean()+(similarity_matrix.mean()*binary_threshold)), 0, 1)
efficiency.append((binay_matrix.sum(axis=0).mean())/len(binay_matrix)*100)
return efficiency
def clustering_function(data:np.array,df,algo:['k_means','hierarchical']='k_means',k=15):
cluster_ind = []
if algo == 'k_means':
clusterer = KMeansClusterer(num_means=k, distance=nltk.cluster.util.cosine_distance, avoid_empty_clusters=True, repeats=1, normalise=True)
assigned_clusters = np.array(clusterer.cluster(data, assign_clusters=True))
else:
clusterer = AgglomerativeClustering(n_clusters=k, affinity="cosine", linkage='complete').fit(data)
assigned_clusters = np.array(clusterer.labels_)
for i in range(k):
cluster_index = np.where(assigned_clusters==i)[0]
cluster_ind.append(cluster_index)
return cluster_ind
def noise_handling(sentence_embeddings, df, cluster_id, binary_threshold=.2, noise_threshold=0.5):
noisy_tickets=[]
pure_cluster_set = []
for i in cluster_id:
if len(i) < Cluster_size :
noisy_tickets.append(i)
else:
one_cluster_index=sentence_embeddings[i]
similarity_matrix=cosine_similarity(one_cluster_index,one_cluster_index)
binary_matrix = np.where(similarity_matrix < binary_threshold, 0, 1)#np.where(similarity_matrix<(similarity_matrix.mean()+(similarity_matrix.mean()*binary_threshold)), 0, 1)
noisy_ticket_index = np.where(binary_matrix.sum(axis=0) <= math.ceil(noise_threshold*len(binary_matrix)))[0]#b.sum(axis=0).max()*threshold)
noisy_tickets_id = np.take(i,noisy_ticket_index) #or use[i[z] for z in noisy_ticket_index[0]]
remaining_tickets_id = np.delete(i,noisy_ticket_index)
remaining_index = np.array([np.argwhere(i == j) for j in remaining_tickets_id]).reshape(-1)
# calculate a cluster efficiency with noise
with_noise_efficiency_score = (binary_matrix.sum(axis=0).mean()) / len(binary_matrix) * 100
if len(noisy_tickets_id) > math.floor(len(i)*0.8):
if len(remaining_index)> Cluster_size:
# calculate a cluster efficiency without noise
binary_matrix_without_noise = np.delete(binary_matrix[remaining_index], noisy_ticket_index, axis=1)
without_noise_binary_matrix_score = (binary_matrix_without_noise.sum(axis=0).mean()) / len(binary_matrix_without_noise) * 100
if without_noise_binary_matrix_score > Pure_cluster_threshold * 100:
pure_cluster_set.append((remaining_tickets_id,without_noise_binary_matrix_score))
else:
noisy_tickets.append((remaining_tickets_id,without_noise_binary_matrix_score))
noisy_tickets.append((noisy_tickets_id,with_noise_efficiency_score))
else:
# calculate a cluster efficiency without noise
binary_matrix_without_noise= np.delete(binary_matrix[remaining_index],noisy_ticket_index,axis=1)
without_noise_binary_matrix_score = (binary_matrix_without_noise.sum(axis=0).mean()) / len(binary_matrix_without_noise) * 100
noisy_tickets.append((noisy_tickets_id,with_noise_efficiency_score))
if without_noise_binary_matrix_score > Pure_cluster_threshold*100:
pure_cluster_set.append((remaining_tickets_id,without_noise_binary_matrix_score))
else:
noisy_tickets.append((remaining_tickets_id,without_noise_binary_matrix_score))
pure_cluster = save_cluster(cluster_set=pure_cluster_set,df=df,path=Cluster_ouput_path,label='pure')
noisy_cluster = save_cluster(cluster_set=noisy_tickets,df=df,path=Cluster_ouput_path,label='noisy')
#cluster_info = show_cluster_information(full_cluster_index=cluster_id, noisy_cluster_index=noisy_tickets, df=df, path=Cluster_ouput_path)
#if Recursion_phase > 1:
# if len(pure_cluster) != 0:
# # Do Cluster again on Noisy Tickets
# sentence_embeddings = sentence_embeddings[list(itertools.chain.from_iterable(noisy_tickets))]
# df = df.iloc[list(itertools.chain.from_iterable(noisy_tickets))]#[df[index] for index in list(itertools.chain.from_iterable(noisy_tickets))]
# k = math.ceil(pow(len(sentence_embeddings), Power_constant))
#
# _,assigned_cluster_id,_ = clustering_engine(data=sentence_embeddings,df=df,file_name=cluster_output_file,algo=algo,k=k)
# #assigned_cluster_id = clustering_function(data=sentence_embeddings, algo=algo, k=k)
#
# result.append(noise_handling(sentence_embeddings=sentence_embeddings, df=df, cluster_id=assigned_cluster_id, binary_threshold=binary_threshold,noise_threshold=noise_threshold))
return pure_cluster,noisy_cluster
def save_cluster(cluster_set,df,path,label:['noisy','pure']='pure'):
counter = 0
data_frames = []
for one_cluster in cluster_set:
cluster = one_cluster[0]
score = one_cluster[1]
cluster_df = df.iloc[cluster]
cluster_df.columns = ['cluster {}'.format(counter)]
cluster_df = cluster_df.append({cluster_df.columns.values[0]: 'Score = {}'.format(score)},ignore_index=True)
data_frames.append(cluster_df)
counter=counter+1
if len(data_frames)>1:
cluster_df = pd.concat(objs=data_frames, axis=1)
else:
if data_frames :
cluster_df = data_frames[0]
else:
return pd.DataFrame()
cluster_df.to_csv(os.path.join(path,label+" cluster_{}.csv".format(counter)))
return cluster_df
def find_nearest(array, value):
array = np.asarray(array)
idx = (np.abs(array - value)).argmin()
return array[idx]
def find_k(sentence_embeddings,df,algo,cluster_scoring_function):
score=[]
Max_k = math.ceil(pow(len(sentence_embeddings),Power_constant))
for k in range(Min_cluster,Max_k):
_,assigned_cluster_id,assigned_cluster = clustering_engine(data=sentence_embeddings,df=df,file_name=cluster_output_file,algo=algo,k=k)
if cluster_scoring_function == 'davies_bouldin_score':
score.append(davies_bouldin_score(sentence_embeddings, assigned_cluster))
elif cluster_scoring_function == 'silhouette_score':
score.append(silhouette_score(sentence_embeddings, assigned_cluster))
elif cluster_scoring_function == 'calinski_harabaz_score':
score.append(calinski_harabaz_score(sentence_embeddings, assigned_cluster))
else:
score.append(base.dunn_fast(sentence_embeddings,assigned_cluster))
#plot_cluster_score(score=score,Min_cluster=Min_cluster,Max_cluster=int(len(df)/Desired_cluster),algo=algo,scoring_function=cluster_scoring_function)
k_index = np.argsort(score)[1:Window_size_for_k]
final_k = find_nearest(k_index,Max_k)
return final_k
def main():
df = pd.read_excel('Tickets.xlsx')
sentences = df.values.flatten().tolist()
preprocessed_sentences = remove_null_sentence(text_preprocessing(sentences=sentences))
input_sentence = word_tokenizer(preprocessed_sentences)
# sentences= []
# with open('data/preprocessed_data.pkl', 'rb') as handle:
# input_sentences = pickle.load(handle)
# for i in input_sentences:
# sentences.append(' '.join(i))
if pretrained_word2vec :
# load word vectors
words, We = data_io.getWordmap(textfile=os.path.join(Word2Vec_path,word2vecfile))
# load word weights
word2weight = data_io.getWordWeight(os.path.join(Word2Vec_path,weightfile),a=weightpara) # word2weight['str'] is the weight for the word 'str'
weight4ind = data_io.getWeight(words, word2weight) # weight4ind[i] is the weight for the i-th word
else:
word2vec_model = train_word2vec(Input_sentence=input_sentence, Embedding_size=100, path=Word2Vec_path)
weightfile_path = save_dict_to_file(word2vec_model=word2vec_model,path=Word2Vec_path)
# load word vectors
words, We = data_io.getWordmap(textfile=os.path.join(Word2Vec_path,word2vecfile))
# load word weights
word2weight = data_io.getWordWeight(weightfile=os.path.join(Word2Vec_path,weightfile),a=weightpara) # word2weight['str'] is the weight for the word 'str'
weight4ind = data_io.getWeight(words, word2weight) # weight4ind[i] is the weight for the i-th word
#load_dict_from_file(path='models/Word2vec/freqency_dict.txt')
if SIF_embedding_generate:
sentence_embeddings = generate_SFI_embedding(sentences=preprocessed_sentences,words=words,We=We,weight4ind=weight4ind,params=params,path=SIF_embedding_path,phase='training')
else:
sentence_embeddings = np.load(os.path.join(SIF_embedding_path,'SIF_embeddings.npy'))
if test:
testing_embedding(total_embedding=sentence_embeddings,words=words,weight4ind=weight4ind,We=We,df=df)
if k_calculation:
k_range = find_k(sentence_embeddings=sentence_embeddings,df=df,algo=algo,cluster_scoring_function=cluster_scoring_function)
else:
k_range = Fixed_K
_, assigned_cluster_id, assigned_cluster = clustering_engine(data=sentence_embeddings, df=df, file_name=cluster_output_file, algo=algo, k=k_range)
noisy_ticket_index, cluster_efficiency, remaining_ticket_index = noise_handling(sentence_embeddings,cluster_id=assigned_cluster_id,df=df,binary_threshold=Binary_threshold,noise_threshold=Noise_threshold)
#cluster_info = show_cluster_information(assigned_cluster_id, noisy_ticket_index,df=df,path=Cluster_ouput_path)
if __name__ == "__main__":
# set parameters
params = params.params()
params.rmpc = rmpc
main()