Skip to content
Mohsen Askar

Visualization Module

Overview

The Visualization module enables you to create interactive and static visualizations of your text data and topic models. It provides a comprehensive suite of visualization tools that help you understand document content, explore word patterns, and interpret topic modeling results through visual representations.

This module allows you to: - Generate word clouds and frequency charts to visualize key terms - Analyze n-gram patterns to identify common phrases - Explore topic distributions across your document collection - Visualize topic keywords and their importance - Highlight topic patterns directly in original texts - Create interactive visualizations for deep exploration of topic models - Export visualizations for presentations and reports

Visualizations make complex text and topic data more accessible, helping to communicate insights effectively and discover patterns that might be missed in raw data.

Main interface of the Visualizer Tab

png png

Core Components

Visualizer Class

The Visualizer class serves as the main interface for visualization functionality, managing the generation of visualizations using worker threads.

Methods

Method Description Parameters Returns
generate_visualization() Starts generation of a visualization in a worker thread viz_type: Type of visualization to generate
data: Data for the visualization
**kwargs: Additional parameters		 Returns the worker thread that can be connected to signals
highlight_topics_in_documents() Highlights topics in original text documents documents: List of document strings
topic_model: Topic model
topic_ids: List of topic IDs to highlight
colors: List of colors for each topic		 List of HTML strings with highlighted topics
get_available_visualizations() Returns available visualization types None Dictionary mapping visualization names to their types
save_figure() Saves a figure to a file fig: Matplotlib figure
filename: Output filename
dpi: Resolution		 Boolean indicating success
figure_to_bytes() Converts a figure to bytes for embedding fig: Matplotlib figure
format: Output format
dpi: Resolution		 Bytes containing the image data

VisualizationWorker Class

The VisualizationWorker class extends QThread to handle visualization generation in a background thread, keeping the UI responsive.

Signals

Signal Description Parameters
progress_updated Emitted to update progress int: progress percentage, str: status message
visualization_completed Emitted when visualization is done object: Figure object or tuple with figure and HTML
error_occurred Emitted when an error occurs str: error message

Visualization Types

The module provides various visualization types for different analytical purposes:

Text-Based Visualizations

Word Cloud

Generates a visual representation of word frequency where word size corresponds to frequency.

Word Frequency

Creates a horizontal bar chart showing the most frequent words in the documents.

N-grams Analysis

Visualizes the most common phrases (word sequences) in the documents.

Topic-Based Visualizations

Topic Distribution

Shows how documents are distributed across topics using a bar chart.

Topic Keywords

Visualizes the most important words for each topic in a grid of bar charts.

Topic Heatmap

Creates a heatmap showing document-topic assignments, where color intensity represents probability.

Topic Highlighting

Highlights topics directly in the original text documents using colored spans.

Interactive Visualizations

pyLDAvis (LDA Visualization)

Creates an interactive visualization for LDA topic models using the pyLDAvis library.

BERTopic Visualization

Creates an interactive visualization for BERTopic models showing topic relationships.

User Interface

The VisualizerTab class provides a comprehensive interface for creating, viewing, and saving visualizations.

UI Components

  • Visualization Controls:
  • Visualization type dropdown
  • Parameter configuration section (changes based on selected visualization)

  • Generate and Save buttons

  • Visualization Display:

  • Tabs for static and interactive views
  • Canvas with toolbar for static visualizations

  • Web view for interactive visualizations

  • Topic Highlighting Dialog:

  • Topic selection with color indication
  • Document navigator for exploring highlighted content
  • Legend showing topic colors and keywords

Usage Guide

Creating Visualizations

  1. Select a visualization type from the dropdown menu
  2. Configure the visualization parameters (if applicable)
  3. Click "Generate Visualization" to create the visualization
  4. The visualization will appear in the main viewing area
  5. Interactive visualizations will appear in the "Interactive View" tab

Customizing Visualizations

Each visualization type offers specific customization options:

Word Cloud

  • Language: Select the text language for better processing
  • Color Scheme: Choose from various color palettes

Word Frequency

  • Number of Words: Set how many top words to display
  • Language: Select the text language

N-grams Analysis

  • N-gram Size: Set the size of word sequences (2=bigrams, 3=trigrams)
  • Number of N-grams: Set how many top phrases to display
  • Language: Select the text language

Topic Keywords

  • Words per Topic: Set how many words to show for each topic
  • Color Scheme: Choose from various color palettes

Topic Highlighting

  • Select Topics: Choose which topics to highlight
  • Color Scheme: Select a color palette for topic differentiation

Interactive LDA

  • MDS Algorithm: Choose between t-SNE and PCoA for dimensionality reduction

BERTopic Visualization

  • Number of Topics: Set how many topics to include in the visualization

The interface provides several tools for exploring visualizations:

  • Zoom and Pan: Use the toolbar below visualizations to zoom in/out and pan
  • Reset View: Click the home button in the toolbar to reset the view
  • Interactive Elements: For interactive visualizations, click on elements for details
  • Smooth Scrolling: Scroll smoothly through large visualizations with optimized controls

Saving Visualizations

  1. Click "Save Visualization" to export your visualization
  2. For static visualizations (word clouds, charts, etc.):
  3. Choose from PNG, JPEG, PDF, or SVG formats
  4. Set the filename and location
  5. For interactive visualizations (pyLDAvis, BERTopic):
  6. Save as HTML for interactive viewing in any web browser
  7. The HTML file contains the complete interactive visualization

Visualization Selection Guide

Visualization Type Best Used For Requirements
Word Cloud Quick overview of key terms, presentations Text documents
Word Frequency Precise word frequency analysis Text documents
N-grams Identifying common phrases and expressions Text documents
Topic Distribution Understanding topic prevalence Topic model results
Topic Keywords Interpreting topic meaning Topic model results
Topic Heatmap Analyzing document-topic relationships Topic model results
Topic Highlighting Validating topics in original context Topic model and documents
Interactive LDA Deep exploration of LDA topic models LDA topic model
BERTopic Visualization Interactive exploration of BERTopic models BERTopic model

Tips for Effective Visualization

General Visualization Tips

  1. Choose the right visualization for your specific analytical goal
  2. Use consistent color schemes across related visualizations
  3. Balance detail and clarity - sometimes less information is more effective
  4. Consider your audience - simpler visualizations for non-technical viewers
  5. Combine multiple visualization types for comprehensive understanding

Word-Based Visualizations

  • Remove common stopwords before visualization for clearer results
  • For technical content, preserve domain-specific terms
  • Compare n-grams with single words to capture multi-word concepts
  • Consider stemming or lemmatization for more concise visualizations

Topic-Based Visualizations

  • Exclude outlier topics (-1) for cleaner visualizations
  • Use topic highlighting to validate model quality in original text
  • Compare topic distribution with topic coherence metrics
  • Use LLM-refined topic labels for more interpretable visualizations

For Presentations

  • Word clouds and simple bar charts are most accessible to general audiences
  • Use consistent color coding across visualizations for topics
  • Include examples of highlighted text to show topics in context
  • Save as high-resolution images (300 DPI or higher) for print materials

Technical Notes

Smooth Scrolling and Performance

The interface includes optimizations for smooth performance:

  • Pixel-based scrolling for smoother navigation
  • Reduced scroll sensitivity for finer control
  • Optimized rendering for large visualizations
  • Hardware acceleration when available

Model Compatibility

  • pyLDAvis works best with LDA models but can display warnings with others
  • BERTopic visualizations require a BERTopic model with embedding data
  • Topic highlighting works with all model types but with varying accuracy

File Formats

When saving visualizations, different formats offer different advantages:

  • PNG: Good general-purpose format, suitable for presentations
  • JPEG: Smaller file size but lower quality for text
  • PDF: Vector format, excellent for publication
  • SVG: Vector format, good for further editing
  • HTML: Required for interactive visualizations