如何保存 nltk FreqDist 图？

Question

I've tried different methods to save my plot but every thing I've tried has turned up with a blank image and I'm not currently out of ideas. Any help with other suggestions that could fix this? The code sample is below.

word_frequency = nltk.FreqDist(merged_lemmatizedTokens) #obtains frequency distribution for each token
print("\nMost frequent top-10 words: ", word_frequency.most_common(10))
word_frequency.plot(10, title='Top 10 Most Common Words in Corpus')
plt.savefig('img_top10_common.png')

Answer 1

I was able to save the NLTK FreqDist plot, when I first initialized a figure object, then called the plot function and finally saved the figure object.

import matplotlib.pyplot as plt
from nltk.probability import FreqDist

fig = plt.figure(figsize = (10,4))
plt.gcf().subplots_adjust(bottom=0.15) # to avoid x-ticks cut-off
fdist = FreqDist(merged_lemmatizedTokens)
fdist.plot(10, cumulative=False)
plt.show()
fig.savefig('freqDist.png', bbox_inches = "tight")

Answer 2

I think you can try the following:

plt.ion()
word_frequency.plot(10, title='Top 10 Most Common Words in Corpus')
plt.savefig('img_top10_common.png')
plt.ioff()
plt.show()

This is because inside nltk's plot function, plt.show() is called and once the figure is closed, plt.savefig() has no active figure to save anymore.

The workaround is to turn interactive mode on, such that the plt.show() from inside the nltk function does not block. Then savefig is called with a current figure available and saves the correct plot. To then show the figure, interactive mode needs to be turned off again and plt.show() be called externally - this time in a blocking mode.

Ideally, nltk would rewrite their plotting function to either allow to set the blocking status, or to not show the plot and return the created figure, or to take a Axes as input to which to plot. Feel free to reach out to them with this request.