-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdisplay_topics.py
219 lines (172 loc) · 9.17 KB
/
display_topics.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
# -*- coding: utf-8 -*-
"""
by mdja, itb, 2019
"""
import numpy as np
import pandas as pd
import topic_utilities
import generate_colormap
#from IPython.display import display
#from tqdm import tqdm
# abstract syntax tree
import ast
import matplotlib.pyplot as plt
#import matplotlib.mlab as mlab
from sklearn.feature_extraction.text import CountVectorizer
#from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.externals import joblib
from textblob import TextBlob
from bokeh.plotting import figure, output_file, show
from bokeh.models import Label
'''
Display most used words in corpus
'''
def display_top_n_words(text_data, n_top_words=10):
count_vectorizer = CountVectorizer(stop_words='english')
words, word_values = topic_utilities.get_top_n_words(n_top_words, count_vectorizer, text_data)
fig, ax = plt.subplots(figsize=(30,8))
ax.bar(range(len(words)), word_values)
ax.set_xticks(range(len(words)))
ax.set_xticklabels(words)
ax.set_title('Top Words')
'''
Generate a histogram of paper word lengths, and use part-of-speech tagging to understand
the types of words used across the corpus. This requires first converting all paper text strings to TextBlobs
and calling the ```pos_tags``` method on each, yielding a list of tagged words for each paper_text.
A complete list of such word tags is available [here](https://www.clips.uantwerpen.be/pages/MBSP-tags).
'''
def pos_tag_word_papers(text_data):
while True:
try:
tagged_paper_text = pd.read_csv('papers-pos-tagged.csv', index_col=0)
word_counts = []
pos_counts = {}
for paper_text in tagged_paper_text[u'tags']:
paper_text = ast.literal_eval(paper_text)
word_counts.append(len(paper_text))
for tag in paper_text:
if tag[1] in pos_counts:
pos_counts[tag[1]] += 1
else:
pos_counts[tag[1]] = 1
except IOError:
tagged_paper_text = [TextBlob(text_data.iloc[i]).pos_tags for i in range(text_data.shape[0])]
tagged_paper_text = pd.DataFrame({'tags':tagged_paper_text})
tagged_paper_text.to_csv('papers-pos-tagged.csv')
continue
break
print('Jumlah total kata dalam paper (tidak termasuk stop words dll):', np.sum(word_counts))
print('Rata-rata jumlah kata dalam paper (tidak termasuk stop words dll): ', np.mean(word_counts))
pos_sorted_types = sorted(pos_counts, key=pos_counts.__getitem__, reverse=True)
pos_sorted_counts = sorted(pos_counts.values(), reverse=True)
fig, ax = plt.subplots(figsize=(18,8))
ax.bar(range(len(pos_counts)), pos_sorted_counts)
ax.set_xticks(range(len(pos_counts)))
ax.set_xticklabels(pos_sorted_types)
ax.set_title('Part-of-Speech Tagging for NIPS Papers Corpus')
ax.set_xlabel('Type of Word')
'''
display topic category vs count
'''
def display_topics_bar(n_topics, title, vectorizer, document_term_matrix, topic_matrix):
keys = topic_utilities.get_keys(topic_matrix)
categories, counts = topic_utilities.keys_to_counts(keys)
print(categories)
print(counts)
top_n_topic = topic_utilities.get_top_n_words_topics(30, n_topics, keys, document_term_matrix, vectorizer)
for i in range(len(top_n_topic)):
print("Topic {}: ".format(i), top_n_topic[i])
top_3_word_topics = topic_utilities.get_top_n_words_topics(3, n_topics, keys, document_term_matrix, vectorizer)
# labels = ['Topic {}: \n'.format(i) + top_3_word_topics[i] for i in range(len(categories))]
labels = ['Topic {}: '.format(i) for i in range(len(categories))]
for i in range(len(top_3_word_topics)):
print("Topic {}: ".format(i), top_3_word_topics[i])
fig, ax = plt.subplots(figsize=(80,8))
ax.bar(categories, counts)
ax.set_xticks(categories)
ax.set_xticklabels(labels)
ax.set_title(title)
'''
display low dimensional projection of topic vector in documents
'''
def display_tsne(output_fn, title, keys, mean_topic_vectors, top_3_topic, tsne_vectors, colormap):
output_file(output_fn)
plot = figure(title=title.format(n_topics), plot_width=800, plot_height=800)
plot.scatter(x=tsne_vectors[:,0], y=tsne_vectors[:,1], color=colormap[keys])
for t in range(len(mean_topic_vectors)):
# print(t)
label = Label(x=mean_topic_vectors[t][0], y=mean_topic_vectors[t][1],
text=top_3_topic[t], text_color=colormap[t])
plot.add_layout(label)
show(plot)
def print_doc_topics(topic_matrix, top_n_topics, nDoc):
msg = "{weight:02d}% in topic {topic_id:02d} which has top words : {text:s}"
for topic_id, weight in enumerate(topic_matrix[nDoc]):
# print(topic_id)
if weight > 0.01 and topic_id < len(top_n_topics):
# text = top_n_topics[topic_id]
print(msg.format(topic_id=topic_id, weight=int(weight * 100.0), text=top_n_topics[topic_id]))
datafile = 'processed_papers.csv'
raw_data = pd.read_csv(datafile)
reindexed_data = raw_data['clean_content']
reindexed_data.index = raw_data['id']
n_topics = 20
############### DISPLAY WORD STATISTICS ###########################
#display_top_n_words(reindexed_data, 10)
#pos_tag_word_papers(reindexed_data)
############### DISPLAY TOPICS ###########################
############### LSA #########################
tfid_vectorizer = joblib.load('tfidf_vectorizer.dat')
document_term_matrix_tfidf = joblib.load('document_term_matrix_tfidf.dat')
lsa_topic_matrix = joblib.load('lsa_topic_matrix.dat')
lsa_model = joblib.load('lsa_model.dat')
title = 'LSA Topic Category Counts'
display_topics_bar(n_topics, title, tfid_vectorizer, document_term_matrix_tfidf,lsa_topic_matrix)
tsne_model = joblib.load('tsne_lsa_model.dat')
tsne_vectors = joblib.load('tsne_lsa_vectors.dat')
lsa_keys = topic_utilities.get_keys(lsa_topic_matrix)
top_3__word_topics = topic_utilities.get_top_n_words_topics(3, n_topics, lsa_keys, document_term_matrix_tfidf, tfid_vectorizer)
mean_topic_vectors = topic_utilities.get_mean_topic_vectors(n_topics, lsa_keys, tsne_vectors)
output_fn = 'ouputlsatsne.html'
title="t-SNE Clustering of {} LSA Topics"
matlib_colormap = generate_colormap.rand_cmap(n_topics, type='bright', first_color_black=False, last_color_black=False, verbose=True)
colormap = generate_colormap.convert_to_bokeh_colormap(matlib_colormap, n_topics)
display_tsne(output_fn, title, lsa_keys, mean_topic_vectors, top_3__word_topics, tsne_vectors, colormap)
################## NMF ######################
count_vectorizer = joblib.load('count_vectorizer_nmf.dat')
document_term_matrix_tfidf_norm = joblib.load('document_term_matrix_tfidf_norm.dat')
nmf_topic_matrix = joblib.load('nmf_topic_matrix.dat')
nmf_model = joblib.load('nmf_model.dat')
title = 'NMF Topic Category Counts'
display_topics_bar(n_topics, title, count_vectorizer, document_term_matrix_tfidf_norm,nmf_topic_matrix)
tsne_model = joblib.load('tsne_nmf_model.dat')
tsne_vectors = joblib.load('tsne_nmf_vectors.dat')
nmf_keys = topic_utilities.get_keys(nmf_topic_matrix)
top_3__word_topics = topic_utilities.get_top_n_words_topics(3, n_topics, nmf_keys, document_term_matrix_tfidf_norm, count_vectorizer)
mean_topic_vectors = topic_utilities.get_mean_topic_vectors(n_topics, nmf_keys, tsne_vectors)
output_fn = 'ouputldatnmf.html'
title="t-SNE Clustering of {} NMF Topics"
matlib_colormap = generate_colormap.rand_cmap(n_topics, type='bright', first_color_black=False, last_color_black=True, verbose=True)
colormap = generate_colormap.convert_to_bokeh_colormap(matlib_colormap, n_topics)
display_tsne(output_fn, title, nmf_keys, mean_topic_vectors, top_3__word_topics, tsne_vectors, colormap)
############### LDA #########################
count_vectorizer = joblib.load('count_vectorizer_lda.dat')
document_term_matrix_count = joblib.load('document_term_matrix_count_lda.dat')
lda_topic_matrix = joblib.load('lda_topic_matrix.dat')
lda_model = joblib.load('lda_model.dat')
title = 'LDA Topic Category Counts'
display_topics_bar(n_topics, title, count_vectorizer, document_term_matrix_count,lda_topic_matrix)
tsne_model = joblib.load('tsne_lda_model.dat')
tsne_vectors = joblib.load('tsne_lda_vectors.dat')
print("LDA : Log Likelihood: ",lda_model.score(document_term_matrix_count))
print("LDA : Perplexity: ", lda_model.perplexity(document_term_matrix_count))
lda_keys = topic_utilities.get_keys(lda_topic_matrix)
top_3__word_topics = topic_utilities.get_top_n_words_topics(3, n_topics, lda_keys, document_term_matrix_count, count_vectorizer)
mean_topic_vectors = topic_utilities.get_mean_topic_vectors(n_topics, lda_keys, tsne_vectors)
output_fn = 'ouputldatsne.html'
title="t-SNE Clustering of {} LDA Topics"
matlib_colormap = generate_colormap.rand_cmap(n_topics, type='bright', first_color_black=False, last_color_black=True, verbose=True)
colormap = generate_colormap.convert_to_bokeh_colormap(matlib_colormap, n_topics)
display_tsne(output_fn, title, lsa_keys, mean_topic_vectors, top_3__word_topics, tsne_vectors, colormap)
top_n_topic = topic_utilities.get_top_n_words_topics(30, n_topics, lda_keys, document_term_matrix_count, count_vectorizer)
print_doc_topics(lda_topic_matrix, top_n_topic, 1)