def createWordCloudForTopic(self, topicNumber):
print('createWordCloudForTopic: {}'.format(topicNumber))
if (topicNumber < 0) or (topicNumber >= self.numberOfTopics):
return False, ''
word_frequencies = self.ldamodel.show_topic(topicNumber, number_of_topic_words)
# Generate a word cloud image
wordFreqDict = dict(word_frequencies)
wordcloud = WordCloud(width=600, height=600).fit_words(wordFreqDict)
wordcloud.background_color = 'white'
fig = plt.figure(figsize=(6,6))
plt.imshow(wordcloud, interpolation='bilinear')
plt.axis("off")
# Save the image to a temp folder to be sent by Flask
filePath = os.path.join('./state/Temp', 'wordcloud.png')
plt.savefig(filePath)
plt.close(fig)
# Return file as a base64 encoded string
with open(filePath, 'rb') as f:
image_read = f.read()
return True, base64.encodestring(image_read).decode("utf-8")
def show_cloud(words):
d = {}
for a, x in words:
d[a] = x
import matplotlib.pyplot as plt
from wordcloud import WordCloud
wordcloud = WordCloud()
wordcloud.prefer_horizontal = 1
wordcloud.width = 800
wordcloud.height = 800
wordcloud.background_color = 'white'
wordcloud.generate_from_frequencies(frequencies=d)
plt.figure(figsize=(8, 8), facecolor=None)
plt.imshow(wordcloud.recolor(color_func=grey_color_func, random_state=3),
interpolation="bilinear")
plt.axis("off")
plt.show()
font_size,
position,
orientation,
random_state=None,
**kwargs):
return "hsl(0, 0%%, %d%%)" % (80 + font_size / 3)
mask = np.array(Image.open(mask_path))
image_color = ImageColorGenerator(mask)
text = open(data_path).read()
#word cloud settings
wc = WordCloud(font_path=content_font_path)
wc.max_words = 500
wc.background_color = "#FFFFFF"
wc.min_font_size = 7
wc.scale = 5
wc.prefer_horizontal = 1
wc.mode = 'RGBA'
wc.mask = mask
#wc creation
wc = wc.generate(text)
default_colors = wc.to_array()
wc.recolor(color_func=image_color, random_state=3)
image = wc.to_image()
#add border
image_size = image.size
def output():
# Load the dictionary
database_dict = pickle.load(open( data_folder_path + 'updated_database_dict.p', 'rb'))
# pull 'ID' (url_string) from input field and store it
url_string = request.args.get('ID')
# pull the number of clusters from the user form
if request.args.get('Nclusters'):
Nclusters = int(request.args.get('Nclusters'))
else:
Nclusters = 3
if len(url_string) < 5:
url_string = 'http://www.nytimes.com/2015/09/20/opinion/sunday/a-toxic-work-world.html'
if url_string not in database_dict.keys():
database_dict = update_database(url_string, database_dict)
headline = database_dict[url_string]['title']
abstract = database_dict[url_string]['abstract']
article_summary = dict(headline = headline, abstract = abstract)
# Generate a wordcloud plot for the article
Nkeywords = len(database_dict[url_string]['keyword_dict'])
word_freq_list = [(entry['value'], Nkeywords - float(entry['rank'])) for entry in database_dict[url_string]['keyword_dict']]
# clean up wordcloud styles
title_wordcloud = WordCloud().generate_from_frequencies(word_freq_list)
title_wordcloud.background_color = 'white'
title_wordcloud.recolor(color_func=custom_color_func)
# Prepare figure for output.html
fig = Figure()
fig.set_facecolor('None')
ax = fig.add_subplot(111)
ax.imshow(title_wordcloud)
ax.set_axis_off()
canvas = FigureCanvas(fig)
title_cloud_png_output = StringIO.StringIO()
canvas.print_png(title_cloud_png_output)
title_cloud_png_output = title_cloud_png_output.getvalue().encode('base64')
# Get the three representative comments
rep_comments = get_representative_comments(database_dict[url_string]['comments_df'], Nclusters)
senti_pos = database_dict[url_string]['comments_df']['senti_pos']
# Get the pie chart
fig = Figure()
fig.set_facecolor('None')
ax = fig.add_subplot(111)
color_repo = ['#4D4D4D', '#5DA5DA', '#FAA43A', '#60BD68', '#F17CB0', '#B2912F', '#B276B2', '#DECF3F', '#F15854']
sizes = [rep_comments[i]['count'] for i in range(Nclusters)]
colors = [color_repo[i] for i in range(Nclusters)]
sorted_sizes_args = np.argsort(sizes)[::-1]
labels = ['Cluster ' + str(i+1) for i in range(Nclusters)]
sorted_sizes = sorted(sizes)[::-1]
ax.pie(sorted_sizes, labels=labels, colors=colors,
autopct='%1.1f%%', shadow=True, startangle=90)
ax.set_axis_off()
canvas = FigureCanvas(fig)
pie_png_output = StringIO.StringIO()
canvas.print_png(pie_png_output)
pie_png_output = pie_png_output.getvalue().encode("base64")
# Sentiment plot
fig = Figure()
fig.set_facecolor('None')
ax = fig.add_subplot(111)
ax.hist(senti_pos, bins = 20)
ax.set_xlim(0,1)
ax.set_xlabel('Sentiment Scale')
ax.set_ylabel('Number of Comments')
ax.set_xticks(np.linspace(0, 1, 5))
senti_labels = [item.get_text() for item in ax.get_xticklabels()]
senti_labels[0] = 'Neg'
senti_labels[2] = '0'
senti_labels[-1] = 'Pos'
ax.set_xticklabels(senti_labels)
canvas=FigureCanvas(fig)
png_output = StringIO.StringIO()
canvas.print_png(png_output)
png_output = png_output.getvalue().encode("base64")
# Return word clouds for different comments
word_cloud_comments = {} # key: cluster_label, val: fig_data
for lab in range(Nclusters):
comment_keywords = get_keywords(rep_comments[lab]['comment'])
# comment_keywords = rep_comments[lab]['cluster_keywords']
keyword_wordfreq_list = [ (word, len(comment_keywords) - i) for i,word in enumerate(comment_keywords)]
wordcloud = WordCloud().generate_from_frequencies(keyword_wordfreq_list)
wordcloud.background_color = 'white'
wordcloud.recolor(color_func=custom_color_func)
fig = Figure()
fig.set_facecolor('None')
ax = fig.add_subplot(111)
ax.imshow(wordcloud)
ax.set_axis_off()
canvas=FigureCanvas(fig)
comment_cloud_png_output = StringIO.StringIO()
canvas.print_png(comment_cloud_png_output)
comment_cloud_png_output = comment_cloud_png_output.getvalue().encode("base64")
word_cloud_comments[lab] = urllib.quote(comment_cloud_png_output.rstrip('\n'))
print word_cloud_comments[0].__class__.__name__
comment_div_str = ''
for i in range(Nclusters):
temp_str = '<div class="row"> <div class="col-md-12" > <h4> Cluster ' + str(i+1) + '</h4> <div class="col-md-5" > <img src="data:image/png;base64,' + word_cloud_comments[sorted_sizes_args[i]] + '" width="400" height="300"/> </div>' + '<div class="col-md-7" > <p class="lead"> ' + rep_comments[sorted_sizes_args[i]]['comment'] + '</p> </div>' + '</div> </div> '
# temp_str = '<div class="row"> <div class="col-md-12" > <h4> Cluster ' + str(i+1) + 'count: ' + str(sizes[sorted_sizes_args[i]]) + '</h4> <div class="col-md-5" > <img src="data:image/png;base64,' + word_cloud_comments[sorted_sizes_args[i]] + '" width="400" height="300"/> </div>' + '<div class="col-md-7" > <p class="lead"> ' + rep_comments[sorted_sizes_args[i]]['comment'] + '</p> </div>' + '</div> </div> '
comment_div_str += temp_str
return render_template("output.html", title_png = urllib.quote(title_cloud_png_output.rstrip('\n')),
article_summary = article_summary, RepComment = rep_comments, comment_coulds = word_cloud_comments,
pie_png = urllib.quote(pie_png_output.rstrip('\n')), img_data=urllib.quote(png_output.rstrip('\n')),
Nclusters = Nclusters, comment_div_str=comment_div_str)
