如何实现NLTK中Bigram FreqDist筛选后相关词对的可视化绘制？

Absolutely! You can easily filter your BigramFreqDist results to focus only on the relevant word pairs, then visualize them directly with NLTK and Matplotlib. Let’s walk through a practical example using text data from NLTK’s built-in corpus (you can swap this out with your own text).

Step 1: Set Up Your Tools #

First, import the necessary libraries and load your text data. We’ll use Moby Dick from nltk.book for this example:

import nltk
from nltk.collocations import BigramCollocationFinder
from nltk.book import text1
import matplotlib.pyplot as plt

# Uncomment below if you haven't downloaded NLTK's book data yet
# nltk.download('book')

Step 2: Generate the Bigram Frequency Distribution #

Create a BigramCollocationFinder to extract bigrams and their frequencies:

# Extract bigrams from the text
bigram_finder = BigramCollocationFinder.from_words(text1)
# Get the frequency distribution of all bigrams
bigram_freq = bigram_finder.ngram_fd

Step 3: Filter Relevant Bigram Pairs #

Choose a filtering strategy that matches your needs—here are two common scenarios:

Option 1: Filter by Frequency (e.g., Only Pairs with Count ≥ X)

If you want to focus on the most frequent bigrams, set a minimum frequency threshold:

# Keep only bigrams that appear 50+ times
filtered_by_freq = [(pair, count) for pair, count in bigram_freq.items() if count >= 50]
# Convert to a new FreqDist object for easy plotting
filtered_freq_dist = nltk.FreqDist(dict(filtered_by_freq))

Option 2: Filter by Keyword (e.g., Pairs Containing a Specific Word)

If you care about bigrams that include a target word (like "whale"), filter for pairs that contain that word:

# Keep only bigrams that include the word "whale"
filtered_by_word = [(pair, count) for pair, count in bigram_freq.items() if "whale" in pair]
filtered_word_freq_dist = nltk.FreqDist(dict(filtered_by_word))

Step 4: Visualize the Filtered Results #

Use NLTK’s built-in plot() method (which leverages Matplotlib) to visualize your filtered bigrams:

# Plot frequency-filtered bigrams
plt.figure(figsize=(12, 6))
filtered_freq_dist.plot(title="Top Bigrams (Frequency ≥ 50) in Moby Dick")
plt.show()

# Plot keyword-filtered bigrams
plt.figure(figsize=(12, 6))
filtered_word_freq_dist.plot(title="Bigrams Containing 'whale' in Moby Dick")
plt.show()

Extra Tips #

• To show only the top N bigrams, add a number to the plot() method (e.g., filtered_freq_dist.plot(20) for the top 20).
• If your filtered list is empty, adjust your threshold (lower the frequency limit or check if your keyword is spelled correctly in the text).
• For custom styling, use Matplotlib’s built-in functions to tweak colors, labels, or layout.

Hope this helps you zero in on exactly the bigram pairs you’re interested in!

内容的提问来源于stack exchange，提问作者glenndino