Hide keyboard shortcuts

Hot-keys on this page

r m x p   toggle line displays

j k   next/prev highlighted chunk

0   (zero) top of page

1   (one) first highlighted chunk

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

#!/usr/bin/env python 

# -*- coding: utf-8 -*- 

 

"""LDA visualization. 

 

This module contains functions for LDA visualization provided by `DARIAH-DE`_. 

 

.. _Gensim: 

https://radimrehurek.com/gensim/index.html 

.. _Mallet: 

http://mallet.cs.umass.edu/ 

.. _DARIAH-DE: 

https://de.dariah.eu 

https://github.com/DARIAH-DE 

""" 

 

__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.pyplot as plt 

from wordcloud import WordCloud 

import numpy as np 

import os 

import pandas as pd 

#import pyLDAvis.gensim 

import regex 

 

 

log = logging.getLogger('visualization') 

log.addHandler(logging.NullHandler()) 

logging.basicConfig(level = logging.WARNING, 

format = '%(levelname)s %(name)s: %(message)s') 

 

class Visualization: 

def __init__(self, lda_model, corpus, dictionary, doc_labels, interactive=False): 

"""Loads Gensim output for further processing. 

 

The output folder should contain ``corpus.mm``, ``corpus.lda``, as well as 

``corpus_doclabels.txt`` (for heatmap) or ``corpus.dict`` (for interactive 

visualization). 

 

Args: 

lda_model: Path to output folder. 

corpus: 

dictionary: 

doc_labels: 

interactive (bool, optional): True if interactive visualization, 

False if heatmap is desired. Defaults to False. 

 

Returns: 

If `interactive == False`: corpus, model, doc_labels. 

If `interactive == True`: corpus, model, dictionary. 

 

Raises: 

OSError: If directory or files not found. 

ValueError: If no matching values found. 

Unexpected error: Everything else. 

""" 

try: 

log.info("Accessing corpus ...") 

self.corpus = corpus 

log.debug("Corpus available.") 

 

log.info("Accessing model ...") 

self.model = lda_model 

log.debug("Model available.") 

 

if interactive == False: 

log.debug(":param: interactive == False.") 

log.info("Accessing doc_labels ...") 

self.doc_labels = doc_labels 

log.debug("doc_labels accessed.") 

 

elif interactive == True: 

log.debug(":param: interactive == True.") 

log.info("Accessing dictionary ...") 

self.dictionary = dictionary 

log.debug("Dictionary available.") 

log.debug("Corpus, model and dictionary available.") 

 

except OSError as err: 

log.error("OS error: {0}".format(err)) 

raise 

except ValueError: 

log.error("Value error: No matching value found.") 

raise 

except: 

import sys 

log.error("Unexpected error:", sys.exc_info()[0]) 

sys.exit(1) 

raise 

 

def make_heatmap(self): 

"""Generates heatmap from LDA model. 

 

The ingested data (e.g. with `load_gensim_output()`) has to be transmitted 

as parameters. 

 

Args: 

corpus: Corpus created by Gensim, e.g. corpus.mm. 

model: LDA model created by Gensim, e.g. corpus.lda. 

doc_labels (list[str]): List of document labels, e.g. corpus_doclabels.txt. 

 

Returns: 

Matplotlib heatmap figure. 

 

ToDo: 

* add colorbar 

* create figure dynamically? 

http://stackoverflow.com/questions/23058560/plotting-dynamic-data-using-matplotlib 

""" 

no_of_topics = self.model.num_topics 

no_of_docs = len(self.doc_labels) 

doc_topic = np.zeros((no_of_docs, no_of_topics)) 

 

log.info("Accessing topic distribution and topic probability ...") 

for doc, i in zip(self.corpus, range(no_of_docs)): 

topic_dist = self.model.__getitem__(doc) 

for topic in topic_dist: # topic_dist is a list of tuples (topic_id, topic_prob) 

doc_topic[i][topic[0]] = topic[1] 

log.debug("Topic distribution and topic probability available.") 

 

log.info("Accessing plot labels ...") 

topic_labels = [] 

for i in range(no_of_topics): 

topic_terms = [x[0] for x in self.model.show_topic(i, topn=3)] # show_topic() returns tuples (word_prob, word) 

topic_labels.append(" ".join(topic_terms)) 

log.debug("%s plot labels available.", len(topic_labels)) 

 

log.info("Creating heatmap figure ...") 

if no_of_docs > 20 or no_of_topics > 20: 

fig = plt.figure(figsize=(20,20)) # if many items, enlarge figure 

else: 

fig = plt.figure() 

ax = fig.add_subplot(1,1,1) 

ax.pcolor(doc_topic, norm=None, cmap='Reds') 

ax.set_yticks(np.arange(doc_topic.shape[0])+1.0) 

ax.set_yticklabels(self.doc_labels) 

ax.set_xticks(np.arange(doc_topic.shape[1])+0.5) 

ax.set_xticklabels(topic_labels, rotation='90') 

ax.invert_yaxis() 

fig.tight_layout() 

self.heatmap_vis = fig 

log.debug("Heatmap figure available.") 

 

def save_heatmap(self, path, filename='heatmap', ext='png', dpi=200): 

"""Saves Matplotlib heatmap figure. 

 

The created visualization (e.g. with `make_heatmap()`) has to be 

transmitted as parameter. 

 

Args: 

heatmap: plt.figure created by ``matplotlib.pyplot``. 

path(str): Path to output folder. Defaults to global variable `path`. 

 

Returns: 

~/out/corpus_heatmap.png 

""" 

log.info("Saving heatmap figure...") 

try: 

if not os.path.exists(path): 

os.makedirs(path) 

self.heatmap_vis.savefig(os.path.join(path, filename + '.' + ext), dpi=dpi) 

log.debug("Heatmap figure available at %s/%s.%s", path, filename, ext) 

except AttributeError: 

log.error("Run make_heatmap() before save_heatmp()") 

raise 

except FileNotFoundError: 

pass 

 

def make_interactive(self): 

"""Generates interactive visualization from LDA model. 

 

The ingested data (e.g. with `load_gensim_output()`) has to be transmitted 

as parameters. 

 

Args: 

corpus: Corpus created by Gensim, e.g. corpus.mm. 

model: LDA model created by Gensim, e.g. corpus.lda. 

dictionary(dict): Dictionary created by Gensim, e.g. corpus.dict. 

 

Returns: 

pyLDAvis visualization. 

""" 

log.info("Accessing model, corpus and dictionary ...") 

self.interactive_vis = pyLDAvis.gensim.prepare(self.model, self.corpus, self.dictionary) 

log.debug("Interactive visualization available.") 

 

def save_interactive(self, path, filename='corpus_interactive'): 

"""Saves interactive visualization. 

The created visualization (e.g. with `make_interactive()`) has to be 

transmitted as parameter. 

 

Args: 

vis: Interactive visualization created by pyLDAvis. 

path(str): Path to output folder. Defaults to global variable `path`. 

 

Returns: 

~/out/corpus_interactive.html 

~/out/corpus_interactive.json 

""" 

try: 

if not os.path.exists(path): 

os.makedirs(path) 

log.info("Saving interactive visualization ...") 

pyLDAvis.save_html(self.interactive_vis, os.path.join(path, 'corpus_interactive.html')) 

pyLDAvis.save_json(self.interactive_vis, os.path.join(path, 'corpus_interactive.json')) 

pyLDAvis.prepared_data_to_html(self.interactive_vis) 

log.debug("Interactive visualization available at %s/corpus_interactive.html and %s/corpus_interactive.json", path, path) 

except AttributeError: 

log.error("Running make_interactive() before save_interactive() ...") 

raise 

except FileNotFoundError: 

pass 

 

def create_doc_topic(corpus, model, doc_labels): 

# Adapted from code by Stefan Pernes 

"""Creates a document-topic data frame. 

 

Args: 

Gensim corpus. 

Gensim model object. 

List of document labels. 

 

Returns: 

 

""" 

no_of_topics = model.num_topics 

no_of_docs = len(doc_labels) 

doc_topic = np.zeros((no_of_docs, no_of_topics)) 

 

for doc, i in zip(corpus, range(no_of_docs)): # use document bow from corpus 

topic_dist = model.__getitem__(doc) # to get topic distribution froom model 

for topic in topic_dist: # topic_dist is a list of tuples 

doc_topic[i][topic[0]] = 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 = doc_labels, columns = topic_labels) 

doc_topic = doc_topic.transpose() 

# TODO: Stupid construction grown out of quick code adaptations: rewrite the first loop to 

# get rid of the necessity to transpose the data frame!!! 

# TODO: 'visualization' is not the proper place for this function! 

 

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. 

 

Args: 

Document-topic data frame. 

 

Returns: 

 

""" 

data_frame = data_frame.transpose().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 

 

# TODO: recode to get rid of transpose in the beginning 

 

 

def plot_doc_topics(doc_topic, document_index): 

"""Plot topic disctribution in a document. 

 

Args: 

Document-topic data frame. 

Index of the document to be shown. 

 

Returns: 

 

""" 

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 

 

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): 

"""Reads Mallet output (topics with words and word weights) into dataframe.""" 

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_wordlewords(word_scores_grouped, number_of_top_words, topic_nr): 

"""Transform Mallet output for wordle generation.""" 

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() 

print(topic_words) 

word_scores = top_topic_word_scores.loc[:,2].tolist() 

print(word_scores) 

wordlewords = "" 

j = 0 

for word in topic_words: 

word = word 

score = word_scores[j] 

j += 1 

wordlewords = wordlewords + ((word + " ") + str(score)) 

return wordlewords 

 

def plot_wordle_from_mallet(word_weights_file, 

topic_nr, 

number_of_top_words, 

outfolder, 

dpi): 

"""Generate wordles from Mallet output, using the wordcloud module.""" 

 

word_scores_grouped = read_mallet_word_weights(word_weights_file) 

text = get_wordlewords(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 = "wordle_tp"+"{:03d}".format(topic_nr) + ".png" 

plt.savefig(outfolder + figure_filename, dpi=dpi) 

return plt