Pfthb

I have the following code that should plot a wordcloud of a given text in matplotlib and converts it to plotly:

from wordcloud import WordCloud, STOPWORDS
import matplotlib.pyplot as plt
import plotly.graph_objs as go
from plotly.offline import download_plotlyjs, init_notebook_mode, plot, iplot
import plotly.tools as tls

# Thanks : https://www.kaggle.com/aashita/word-clouds-of-various-shapes ##
def plot_wordcloud(text, mask=None, max_words=200, max_font_size=100, figure_size=(24.0,16.0), 
 title = None, title_size=40, image_color=False):
 stopwords = set(STOPWORDS)
 wordcloud = WordCloud(background_color='black',
 stopwords = stopwords,
 max_words = max_words,
 max_font_size = max_font_size, 
 random_state = 42,
 width=800, 
 height=400,
 mask = mask)
 wordcloud.generate(str(text))

 fig = plt.figure()
 plt.imshow(wordcloud)
 return tls.mpl_to_plotly(fig)

word_list = "Wikipedia was launched on January 15, 2001, by Jimmy Wales and Larry Sanger.[10] Sanger coined its name,[11][12] as a portmanteau of wiki[notes 3] and 'encyclopedia'. Initially an English-language encyclopedia, versions in other languages were quickly developed. With 5,748,461 articles,[notes 4] the English Wikipedia is the largest of the more than 290 Wikipedia encyclopedias. Overall, Wikipedia comprises more than 40 million articles in 301 different languages[14] and by February 2014 it had reached 18 billion page views and nearly 500 million unique visitors per month.[15] In 2005, Nature published a peer review comparing 42 science articles from Encyclopædia Britannica and Wikipedia and found that Wikipedia's level of accuracy approached that of Britannica.[16] Time magazine stated that the open-door policy of allowing anyone to edit had made Wikipedia the biggest and possibly the best encyclopedia in the world and it was testament to the vision of Jimmy Wales.[17] Wikipedia has been criticized for exhibiting systemic bias, for presenting a mixture of 'truths, half truths, and some falsehoods',[18] and for being subject to manipulation and spin in controversial topics.[19] In 2017, Facebook announced that it would help readers detect fake news by suitable links to Wikipedia articles. YouTube announced a similar plan in 2018."

plot_wordcloud(word_list, title="Word Cloud")

This just returns a blank figure with nothing in the data part:

Figure(
 'data': ,
 'layout': 'autosize': False,
 'height': 288,
 'hovermode': 'closest',
 'margin': 'b': 61, 'l': 54, 'pad': 0, 'r': 43, 't': 59,
 'showlegend': False,
 'width': 432,
 'xaxis': 'anchor': 'y',
 'domain': [0.0, 1.0],
 'mirror': 'ticks',
 'nticks': 10,
 'range': [-0.5, 799.5],
 'showgrid': False,
 'showline': True,
 'side': 'bottom',
 'tickfont': 'size': 10.0,
 'ticks': 'inside',
 'type': 'linear',
 'zeroline': False,
 'yaxis': 'anchor': 'x',
 'domain': [0.0, 1.0],
 'mirror': 'ticks',
 'nticks': 10,
 'range': [399.5, -0.5],
 'showgrid': False,
 'showline': True,
 'side': 'left',
 'tickfont': 'size': 10.0,
 'ticks': 'inside',
 'type': 'linear',
 'zeroline': False
)

Why is that? And how do I fix it?

If I want to plot the matplotlib plot, it works fine - return fig returns a static figure of the wordcloud.

I tried to directly plot the wordcloud in plotly, but with go.Scatter you need to supply the x and y values explicitly - it cannot take them from wordcloud implicitly like plt.imshow can. So, I get a "object is not iterable" error:

def plot_wordcloud(text, mask=None, max_words=200, max_font_size=100, figure_size=(24.0,16.0), 
 title = None, title_size=40, image_color=False):
 stopwords = set(STOPWORDS)
 wordcloud = WordCloud(background_color='black',
 stopwords = stopwords,
 max_words = max_words,
 max_font_size = max_font_size, 
 random_state = 42,
 width=800, 
 height=400,
 mask = mask)
 wordcloud.generate(str(text))


 data = go.Scatter(dict(wordcloud.generate(str(text))),
 mode='text',
 text=words,
 marker='opacity': 0.3,
 textfont='size': weights,
 'color': colors)
 layout = go.Layout('xaxis': 'showgrid': False, 'showticklabels': False, 'zeroline': False,
 'yaxis': 'showgrid': False, 'showticklabels': False, 'zeroline': False)
 fig = go.Figure(data=[data], layout=layout)
 return fig


word_list = "Wikipedia was launched on January 15, 2001, by Jimmy Wales and Larry Sanger.[10] Sanger coined its name,[11][12] as a portmanteau of wiki[notes 3] and 'encyclopedia'. Initially an English-language encyclopedia, versions in other languages were quickly developed. With 5,748,461 articles,[notes 4] the English Wikipedia is the largest of the more than 290 Wikipedia encyclopedias. Overall, Wikipedia comprises more than 40 million articles in 301 different languages[14] and by February 2014 it had reached 18 billion page views and nearly 500 million unique visitors per month.[15] In 2005, Nature published a peer review comparing 42 science articles from Encyclopædia Britannica and Wikipedia and found that Wikipedia's level of accuracy approached that of Britannica.[16] Time magazine stated that the open-door policy of allowing anyone to edit had made Wikipedia the biggest and possibly the best encyclopedia in the world and it was testament to the vision of Jimmy Wales.[17] Wikipedia has been criticized for exhibiting systemic bias, for presenting a mixture of 'truths, half truths, and some falsehoods',[18] and for being subject to manipulation and spin in controversial topics.[19] In 2017, Facebook announced that it would help readers detect fake news by suitable links to Wikipedia articles. YouTube announced a similar plan in 2018."

plot_wordcloud(word_list, title="Word Cloud")

---------------------------------------------------------------------------

TypeError Traceback (most recent call last)
<ipython-input-50-0567281b72b3> in <module>()

---> 48 plot_wordcloud(word_list, title="Word Cloud")

<ipython-input-50-0567281b72b3> in plot_wordcloud(text, mask, max_words, max_font_size, figure_size, title, title_size, image_color)
 18 
 19 
---> 20 data = go.Scatter(dict(wordcloud.generate(str(text))),
 21 mode='text',
 22 text=words,

TypeError: 'WordCloud' object is not iterable

If I return wordcloud, it displays this: <wordcloud.wordcloud.WordCloud at 0x1c8faeda748>. If anyone knows how to unpack the wordcloud object so that I can input the x and y parameters from it into go.Figure, that would be great as well (better in fact).

Just to show that unpacking the wordcloud object would work, I can natively plot a wordcloud with plotly by putting random numbers for the x and y values in go.Scatter like so:

import random
import plotly.graph_objs as go

def plot_wordcloud(text, mask=None, max_words=200, max_font_size=100, figure_size=(24.0,16.0), 
 title = None, title_size=40, image_color=False):
 stopwords = set(STOPWORDS)
 wordcloud = WordCloud(background_color='black',
 stopwords = stopwords,
 max_words = max_words,
 max_font_size = max_font_size, 
 random_state = 42,
 width=800, 
 height=400,
 mask = mask)
 wordcloud.generate(str(text))


 data = go.Scatter(x=[random.random() for i in range(3000)],
 y=[random.random() for i in range(3000)],
 mode='text',
 text=str(word_list).split(),
 marker='opacity': 0.3,
 textfont='size': weights,
 'color': colors)
 layout = go.Layout('xaxis': 'showgrid': False, 'showticklabels': False, 'zeroline': False,
 'yaxis': 'showgrid': False, 'showticklabels': False, 'zeroline': False)
 fig = go.Figure(data=[data], layout=layout)
 return fig

Its just not the correct wordcloud (obviously, with the positions and sizes of the words correctly defined), which should look like this (the static wordcloud plotted with matplotlib.pyplot):

Since wordcloud produces an image, and plotly's conversion function cannot currently handle images, you would need to somehow regenerate the wordcloud from the positions, sizes and orientations of the wordcloud.wordcloud.WordCloud object.

Those information are stored in the .layout_ attribute

wc = Wordcloud(...)
wc.generate(text)
print(wc.layout_)

prints a list of tuples of the form

[(word, freq), fontsize, position, orientation, color]

e.g. in this case

[(('Wikipedia', 1.0), 100, (8, 7), None, 'rgb(56, 89, 140)'), 
 (('articles', 0.4444444444444444), 72, (269, 310), None, 'rgb(58, 186, 118)'), ...]

So in principle this allows to regenerate the wordcloud as text. However care must be taken for the little details. I.e. the font and fontsize need to be the same.

Here is a pure matplotlib example, which reproduces the wordcloud with matplotlib.text.Text objects.

import numpy as np
from wordcloud import WordCloud, STOPWORDS 
from wordcloud.wordcloud import FONT_PATH
import matplotlib.pyplot as plt
from matplotlib.font_manager import FontProperties

word_list = "Wikipedia was launched on January 15, 2001, by Jimmy Wales and Larry Sanger.[10] Sanger coined its name,[11][12] as a portmanteau of wiki[notes 3] and 'encyclopedia'. Initially an English-language encyclopedia, versions in other languages were quickly developed. With 5,748,461 articles,[notes 4] the English Wikipedia is the largest of the more than 290 Wikipedia encyclopedias. Overall, Wikipedia comprises more than 40 million articles in 301 different languages[14] and by February 2014 it had reached 18 billion page views and nearly 500 million unique visitors per month.[15] In 2005, Nature published a peer review comparing 42 science articles from Encyclopædia Britannica and Wikipedia and found that Wikipedia's level of accuracy approached that of Britannica.[16] Time magazine stated that the open-door policy of allowing anyone to edit had made Wikipedia the biggest and possibly the best encyclopedia in the world and it was testament to the vision of Jimmy Wales.[17] Wikipedia has been criticized for exhibiting systemic bias, for presenting a mixture of 'truths, half truths, and some falsehoods',[18] and for being subject to manipulation and spin in controversial topics.[19] In 2017, Facebook announced that it would help readers detect fake news by suitable links to Wikipedia articles. YouTube announced a similar plan in 2018."

def get_wordcloud(width, height):
 wc = WordCloud(background_color='black',
 stopwords = set(STOPWORDS),
 max_words = 200,
 max_font_size = 100, 
 random_state = 42,
 width=int(width), 
 height=int(height),
 mask = None)
 wc.generate(word_list)
 return wc


fig, (ax, ax2) = plt.subplots(nrows=2, sharex=True, sharey=True)

fp=FontProperties(fname=FONT_PATH)
bbox = ax.get_position().transformed(fig.transFigure)
wc = get_wordcloud(bbox.width, bbox.height)

ax.imshow(wc)

ax2.set_facecolor("black")
for (word, freq), fontsize, position, orientation, color in wc.layout_:
 color = np.array(color[4:-1].split(", ")).astype(float)/255.
 x,y = position
 rot = None : 0, 2: 90[orientation]
 fp.set_size(fontsize*72./fig.dpi)
 ax2.text(y,x, word, va="top", ha="left", color=color, rotation=rot, 
 fontproperties=fp)

print(wc.layout_)
plt.show()

The upper plot is the wordcloud image shown via imshow, the lower plot is the regenerated wordcloud.

Now you might want to do the same in plotly instead of matplotlib, but I'm not profilient enough with plotly to directly give a solution here.