如何基于文档词汇共现矩阵绘制热力图？

This is a perfect use case for a heatmap—it’ll make those co-occurrence patterns way easier to spot at a glance. Let’s walk through how to adjust your code to build the matrix and plot it properly.

Step 1: Finish Building the Co-Occurrence Matrix #

First, let’s fix and complete your code to generate a square matrix where each cell (i,j) represents how many documents contain both names[i] and names[j]. Here’s the corrected version:

import matplotlib.pyplot as plt
import numpy as np

# Load your documents
a = (open('A.txt').read()).lower().split()
c = (open('C.txt').read()).lower().split()
f = (open('F.txt').read()).lower().split()
documents = [a, f, c]

# Replace with your actual list of important words
names = ["word1", "word2", "word3", "word4"]  # Example list

# Initialize an empty co-occurrence matrix
num_words = len(names)
co_occurrence_matrix = np.zeros((num_words, num_words), dtype=int)

# Calculate co-occurrences
for i, word_a in enumerate(names):
    for j, word_b in enumerate(names):
        count = 0
        for doc in documents:
            # Check if both words are present in the document
            if word_a in doc and word_b in doc:
                count += 1
        co_occurrence_matrix[i][j] = count

Note: If you don’t want to count a word co-occurring with itself (the diagonal cells), add a condition if word_a != word_b inside the document loop. Keeping the diagonal can still be useful, though—it shows how often each word appears across all documents.

Step 2: Plot the Heatmap with Seaborn #

Seaborn makes heatmaps straightforward and customizable. First install it if you haven’t (pip install seaborn), then add this code:

import seaborn as sns

# Set up a readable figure size
plt.figure(figsize=(10, 8))

# Create the heatmap
sns.heatmap(co_occurrence_matrix, 
            annot=True,  # Show exact count values in each cell
            fmt='d',  # Format annotations as integers
            cmap='YlGnBu',  # Choose a color palette (adjust as needed)
            xticklabels=names, 
            yticklabels=names)

# Add context with labels and title
plt.title('Document Word Co-Occurrence Heatmap', fontsize=14)
plt.xlabel('Words', fontsize=12)
plt.ylabel('Words', fontsize=12)

# Show the final plot
plt.show()

Quick Tweaks to Improve the Visualization #

• Normalize Counts: If you want to account for how often each word appears individually (instead of raw counts), divide each cell by the total occurrences of either word (or their average) to get a normalized co-occurrence score.
• Adjust Color Palette: Try coolwarm, viridis, or RdBu if you prefer a different color scheme.
• Rotate Labels: For long word lists, add plt.xticks(rotation=45) to prevent label overlap.

This will give you a clear heatmap where more intense colors indicate higher co-occurrence between pairs of words.

内容的提问来源于stack exchange，提问作者Oumnia Amal Rizki