forked from pielstroem/Topics
-
Notifications
You must be signed in to change notification settings - Fork 13
/
visualization.py
executable file
·458 lines (348 loc) · 14.6 KB
/
visualization.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
453
454
# -*- coding: utf-8 -*-
"""
Visualizing the Output of LDA Models
====================================
"""
__author__ = "DARIAH-DE"
__authors__ = "Steffen Pielstroem, Sina Bock, Severin Simmler"
__email__ = "pielstroem@biozentrum.uni-wuerzburg.de"
__version__ = "0.1"
__date__ = "2017-01-20"
import logging
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import numpy as np
import os
import pandas as pd
import regex
from collections import defaultdict
log = logging.getLogger('visualization')
log.addHandler(logging.NullHandler())
logging.basicConfig(level = logging.ERROR,
format = '%(levelname)s %(name)s: %(message)s')
def create_doc_topic(corpus, model, doc_labels):
# Adapted from code by Stefan Pernes
"""Creates a document-topic-matrix.
Description:
With this function you can create a doc-topic-maxtrix for gensim
output.
Args:
corpus (mmCorpus): Gensim corpus.
model: Gensim LDA model
doc_labels (list): List of document labels.
Returns:
Doc_topic-matrix as DataFrame
ToDo:
Example:
>>> import gensim
>>> corpus = [[(1, 0.5)], []]
>>> gensim.corpora.MmCorpus.serialize('/tmp/corpus.mm', corpus)
>>> mm = gensim.corpora.MmCorpus('/tmp/corpus.mm')
>>> type2id = {0 : "test", 1 : "corpus"}
>>> doc_labels = ['doc1', 'doc2']
>>> model = gensim.models.LdaModel(corpus=mm, id2word=type2id, num_topics=1)
>>> doc_topic = visualization.create_doc_topic(corpus, model, doc_labels)
>>> len(doc_topic.T) == 2
>>> True
"""
no_of_topics = model.num_topics
no_of_docs = len(doc_labels)
doc_topic = np.zeros((no_of_topics, no_of_docs))
for doc, i in zip(corpus, range(no_of_docs)): # use document bow from corpus
topic_dist = model.__getitem__(doc) # to get topic distribution from model
for topic in topic_dist: # topic_dist is a list of tuples
doc_topic[topic[0]][i] = topic[1] # save topic probability
topic_labels = []
for i in range(no_of_topics):
topic_terms = [x[0] for x in model.show_topic(i, topn=3)] # show_topic() returns tuples (word_prob, word)
topic_labels.append(" ".join(topic_terms))
doc_topic = pd.DataFrame(doc_topic, index = topic_labels, columns = doc_labels)
return doc_topic
def doc_topic_heatmap(data_frame):
# Adapted from code by Stefan Pernes and Allen Riddell
"""Plot documnet-topic distribution in a heat map.
Description:
Use create_doc_topic() to generate a doc-topic
Args:
data_frame (DataFrame): Document-topic-matrix.
Returns:
Plot with Heatmap
ToDo:
Example:
>>> import gensim
>>> corpus = [[(1, 0.5)], []]
>>> gensim.corpora.MmCorpus.serialize('/tmp/corpus.mm', corpus)
>>> mm = gensim.corpora.MmCorpus('/tmp/corpus.mm')
>>> type2id = {0 : "test", 1 : "corpus"}
>>> doc_labels = ['doc1', 'doc2']
>>> model = gensim.models.LdaModel(corpus=mm, id2word=type2id, num_topics=1)
>>> doc_topic = visualization.create_doc_topic(corpus, model, doc_labels)
>>> plot = doc_topic_heatmap(doc_topic)
>>> plot.get_fignumns()
[1]
"""
data_frame = data_frame.sort_index()
doc_labels = list(data_frame.index)
topic_labels = list(data_frame)
if len(doc_labels) > 20 or len(topic_labels) > 20: plt.figure(figsize=(20,20)) # if many items, enlarge figure
plt.pcolor(data_frame, norm=None, cmap='Reds')
plt.yticks(np.arange(data_frame.shape[0])+1.0, doc_labels)
plt.xticks(np.arange(data_frame.shape[1])+0.5, topic_labels, rotation='90')
plt.gca().invert_yaxis()
plt.tight_layout()
#plt.savefig(path+"/"+corpusname+"_heatmap.png") #, dpi=80)
return plt
def plot_doc_topics(doc_topic, document_index):
"""Plot topic disctribution in a document.
Description:
Args:
Document-topic data frame.
Index of the document to be shown.
Returns:
Plot.
"""
data = doc_topic[list(doc_topic)[document_index]].copy()
data = data[data != 0]
data = data.sort_values()
values = list(data)
labels = list(data.index)
plt.barh(range(len(values)), values, align = 'center', alpha=0.5)
plt.yticks(range(len(values)), labels)
plt.title(list(doc_topic)[document_index])
plt.xlabel('Proportion')
plt.ylabel('Topic')
plt.tight_layout()
return plt
try:
from wordcloud import WordCloud
#
# Work in progress following
#
def topicwords_in_df(model):
pattern = regex.compile(r'\p{L}+\p{P}?\p{L}+')
topics = []
index = []
for n, topic in enumerate(model.show_topics()):
topics.append(pattern.findall(topic[1]))
index.append("Topic " + str(n+1))
df = pd.DataFrame(topics, index=index, columns=["Key " + str(x+1) for x in range(len(topics))])
return df
def show_wordle_for_topic(model, topic_nr, words):
"""Plot wordle for a specific topic
Args:
model: Gensim LDA model
topic_nr(int): Choose topic
words (int): Number of words to show
Note: Function does use wordcloud package -> https://pypi.python.org/pypi/wordcloud
pip install wordcloud
ToDo: Check if this function should be implemented
"""
plt.figure()
plt.imshow(WordCloud().fit_words(dict(model.show_topic(topic_nr, words))))
plt.axis("off")
plt.title("Topic #" + str(topic_nr + 1))
return plt
def get_color_scale(word, font_size, position, orientation, font_path, random_state=None):
""" Create color scheme for wordle."""
return "hsl(245, 58%, 25%)" # Default. Uniform dark blue.
#return "hsl(0, 00%, %d%%)" % random.randint(80, 100) # Greys for black background.
#return "hsl(221, 65%%, %d%%)" % random.randint(30, 35) # Dark blues for white background
def get_topicRank(topic, topicRanksFile):
#print("getting topic rank.")
with open(topicRanksFile, "r") as infile:
topicRanks = pd.read_csv(infile, sep=",", index_col=0)
rank = int(topicRanks.iloc[topic]["Rank"])
return rank
def read_mallet_word_weights(word_weights_file):
"""Read Mallet word_weigths file
Description:
Reads Mallet word_weigths into pandas DataFrame.
Args:
word_weigts_file: Word_weights_file created with Mallet
Returns: Pandas DataFrame
Note:
ToDo:
"""
word_scores = pd.read_table(word_weights_file, header=None, sep="\t")
word_scores = word_scores.sort(columns=[0,2], axis=0, ascending=[True, False])
word_scores_grouped = word_scores.groupby(0)
return word_scores_grouped
def _get_wordcloudwords(word_scores_grouped, number_of_top_words, topic_nr):
"""Transform Mallet output for wordcloud generation.
Description:
Get words for wordcloud.
Args:
word_scores_grouped(DataFrame): Uses read_mallet_word_weights() to get
grouped word scores.
topic_nr(int): Topic the wordcloud should be generated for
number_of_top_words(int): Number of top words that should be considered
Returns: Words for wordcloud.
Note:
ToDo:
"""
topic_word_scores = word_scores_grouped.get_group(topic_nr)
top_topic_word_scores = topic_word_scores.iloc[0:number_of_top_words]
topic_words = top_topic_word_scores.loc[:,1].tolist()
#word_scores = top_topic_word_scores.loc[:,2].tolist()
wordcloudwords = ""
j = 0
for word in topic_words:
word = word
#score = word_scores[j]
j += 1
wordcloudwords = wordcloudwords + ((word + " "))
return wordcloudwords
def plot_wordcloud_from_mallet(word_weights_file,
topic_nr,
number_of_top_words,
outfolder,
dpi):
"""Generate wordclouds from Mallet output.
Description:
This function does use the wordcloud module to plot wordclouds.
Uses read_mallet_word_weigths() and get_wordlewords() to get
word_scores and words for wordclouds.
Args:
word_weigts_file: Word_weights_file created with Mallet
topic_nr(int): Topic the wordclouds should be generated for
number_of_top_words(int): Number of top words that should be considered
for the wordclouds
outfolder(str): Specify path to safe wordclouds.
dpi(int): Set resolution for wordclouds.
Returns: Plot
Note:
ToDo:
"""
word_scores_grouped = read_mallet_word_weights(word_weights_file)
text = _get_wordcloudwords(word_scores_grouped, number_of_top_words, topic_nr)
wordcloud = WordCloud(width=600, height=400, background_color="white", margin=4).generate(text)
default_colors = wordcloud.to_array()
figure_title = "topic "+ str(topic_nr)
plt.imshow(default_colors)
plt.imshow(wordcloud)
plt.title(figure_title, fontsize=30)
plt.axis("off")
## Saving the image file.
if not os.path.exists(outfolder):
os.makedirs(outfolder)
figure_filename = "wordcloud_tp"+"{:03d}".format(topic_nr) + ".png"
plt.savefig(outfolder + figure_filename, dpi=dpi)
return plt
def plot_wordle_from_lda(model, vocab, topic_nr, words, height, width):
topic_dist = model.topic_word_[topic_nr]
topic_words = np.array(vocab)[np.argsort(topic_dist)][:-words:-1]
token_value = {}
for token, value in zip(topic_words, topic_dist[:-words:-1]):
token_value.update({token: value})
return WordCloud(background_color='white', height=height, width=width).fit_words(token_value)
except ImportError as e:
log.info('WordCloud functions not available, they require the wordcloud module')
def doc_topic_heatmap_interactive(doc_topic, title):
"""Plot interactive doc_topic_heatmap
Description:
With this function you can plot an interactive doc_topic matrix.
Args:
doc_topic (DataFrame): Doc_topic matrix in a DataFrame
title (str): Title shown in the plot.
Returns: bokeh plot
Note:
ToDo:
"""
log.info("Importing functions from bokeh ...")
try:
#from ipywidgets import interact
from bokeh.io import output_notebook
from bokeh.plotting import figure
from math import pi
from bokeh.models import (
ColumnDataSource,
HoverTool,
LinearColorMapper,
BasicTicker,
ColorBar
)
output_notebook()
documents = list(doc_topic.columns)
topics = doc_topic.index
score = []
for x in doc_topic.apply(tuple):
score.extend(x)
data = {
'Topic': list(doc_topic.index) * len(doc_topic.columns),
'Document': [item for item in list(doc_topic.columns) for i in range(len(doc_topic.index))],
'Score': score
}
df = doc_topic.from_dict(data)
colors = ["#c6dbef", "#9ecae1", "#6baed6", "#4292c6", "#2171b5", "#08519c", "#08306b"]
mapper = LinearColorMapper(palette=colors, low=df.Score.min(), high=df.Score.max())
source = ColumnDataSource(df)
TOOLS = "hover,save,pan,box_zoom,reset,wheel_zoom"
p = figure(title=title,
x_range=documents, y_range=list(reversed(topics)),
x_axis_location="above", plot_width=1024, plot_height=768,
tools=TOOLS, toolbar_location='below', responsive=True)
p.grid.grid_line_color = None
p.axis.axis_line_color = None
p.axis.major_tick_line_color = None
p.axis.major_label_text_font_size = "9pt"
p.axis.major_label_standoff = 0
p.xaxis.major_label_orientation = pi / 3
p.rect(x="Document", y="Topic", width=1, height=1,
source=source,
fill_color={'field': 'Score', 'transform': mapper},
line_color=None)
color_bar = ColorBar(color_mapper=mapper, major_label_text_font_size="10pt",
ticker=BasicTicker(desired_num_ticks=len(colors)),
label_standoff=6, border_line_color=None, location=(0, 0))
p.add_layout(color_bar, 'right')
p.select_one(HoverTool).tooltips = [
('Document', '@Document'),
('Topic', '@Topic'),
('Score', '@Score')
]
return p
except:
log.info("Bokeh could not be imported now using mathplotlib")
doc_topic_heatmap(doc_topic)
p.add_layout(color_bar, 'right')
p.select_one(HoverTool).tooltips = [
('Document', '@Document'),
('Topic', '@Topic'),
('Score', '@Score')
]
return p
def show_topic_over_time(doc_topic, labels=['armee truppen general', 'regierung preußen partei', 'dichter goethe kunst'], threshold=0.1, starttime=1841, endtime=1920):
"""Creates a visualization that shows topics over time
Description:
With this function you can plot topics over time using metadata stored in the documents name.
Only works with mallet output.
Args:
doc_topic: doc-topic matrix created by mallet.show_doc_topic_matrix
labels(list[str]): first three keys in a topic to select
threshold(float): threshold set to define if a topic in a document is viable
starttime(int): sets starting point for visualization
endtime(int): sets ending point for visualization
Returns: matplotlib plot
Note: this function is created for a corpus with filenames that looks like:
1866_ArticleName.txt
ToDo: make it compatible with gensim output
"""
years=list(range(starttime,endtime))
doc_topicT = doc_topic.T
for label in labels:
topic_over_threshold_per_year =[]
df = doc_topicT.loc[doc_topicT[label] > threshold]
d = defaultdict(int)
for item in df.index.values:
year = item.split('_')
d[year[0]]+=1
for year in years:
topic_over_threshold_per_year.append(d[str(year)])
plt.plot(years, topic_over_threshold_per_year, label=label)
plt.xlabel('Year')
plt.ylabel('count topics over threshold')
plt.legend()
fig = plt.gcf()
fig.set_size_inches(18.5, 10.5)
plt.show()