Complete guide to using ConGraCNet for linguistic research and network analysis
- Getting Started
- Application Interface
- Corpus Selection
- Lemma and POS Selection
- Network Construction
- Network Analysis
- Sentiment Analysis
- Visualization Options
- Advanced Features
- Troubleshooting
-
Start the Application
streamlit run cgcnStream_0_3_6_withSBBLabel.py
-
Access the Web Interface
- Open your browser to
http://localhost:8501 - The application will load with default settings
- Open your browser to
-
Verify Database Connection
- Check the sidebar for connection status
- Ensure Neo4j database is accessible
The ConGraCNet interface is organized into several main sections:
- Header: Application title and version information
- Main Content Area: Dynamic content based on current operation
- Sidebar: Configuration options and parameters
- Tabs: Organized feature sections (Semantic Dictionaries, Sentiment Analysis, etc.)
- Application Title: "ConGraCNet"
- Version Information: Current version (3.6.2)
- Introduction: Markdown content explaining the application
- Corpus Dropdown: Select from available corpora
- Language Support: Croatian (hr) and English (en)
- API Integration: Automatic data fetching from Sketch Engine
- Lemma Input: Text field for entering base word forms
- POS Selection: Part-of-speech specification
- Combined Form: Automatic combination (lemma+pos)
- Network Parameters: Core network construction settings
- Visualization Options: Display and layout preferences
- Analysis Settings: Sentiment and clustering parameters
ConGraCNet supports multiple corpora with different characteristics:
- hrWaC22: Croatian Web Corpus 2022
- hrTenTen13: Croatian TenTen Corpus 2013
- hrWaC: Croatian Web Corpus
- enTenTen13: English TenTen Corpus 2013
- enWaC: English Web Corpus
-
Open Corpus Dropdown
- Located at the top of the main content area
- Shows all available corpora for your account
-
Choose Appropriate Corpus
- Consider language requirements
- Select based on domain relevance
- Consider corpus size and coverage
-
Verify Selection
- Check that corpus information loads correctly
- Verify language detection
- Confirm grammatical relation availability
After selection, the system displays:
- Corpus Name: Full corpus identifier
- Language: Primary language (hr/en)
- Corpus ID: Unique identifier for database queries
- Initial GramRel: Default grammatical relation type
- Initial Lexeme: Suggested starting word
- Initial POS: Suggested part of speech
-
Enter Base Form
- Use the dictionary form of the word
- Avoid inflected forms (e.g., "run" not "running")
- Ensure correct spelling
-
Examples
- Croatian: "kuca" (house), "auto" (car), "grad" (city)
- English: "house", "car", "city", "computer"
-
POS Categories
- N: Noun
- V: Verb
- ADJ: Adjective
- ADV: Adverb
- PREP: Preposition
-
Selection Process
- Choose from dropdown menu
- System automatically adds "-" prefix
- Combined form: lemma + "-" + pos
-
Validation
- System checks if lemma+pos exists in corpus
- Error message if combination not found
- Suggests alternatives if available
The system automatically creates a combined form:
- Format:
lemma-pos - Example: "kuca-N" (house-noun)
- Usage: Used throughout the application for queries
- Purpose: Number of co-occurring words to analyze
- Range: 1-100 (default: 15)
- Impact: Larger values create more complex networks
- Consideration: Balance between detail and performance
- Score: Association strength (recommended for analysis)
- Frequency: Raw occurrence count
- Usage: Choose based on analysis goals
- Purpose: Fetch initial co-occurrence data
- When to Use: First time analyzing a lemma
- Process: Automatic API calls to Sketch Engine
- Status: Shows progress in spinner
- Purpose: Fetch friend-of-friend data
- When to Use: Building second-degree networks
- Impact: Expands network coverage
- Consideration: Increases processing time
- Definition: Relationship between source lemma and friends
- Selection: Choose from available relations
- Examples:
- Croatian: "obj" (object), "subj" (subject), "mod" (modifier)
- English: "obj", "subj", "mod", "prep"
- Definition: Relationship between friends and their friends
- Selection: Choose from available relations
- Process: System calculates available options
- Impact: Determines network depth
Source Lemma (gr1) → Friends (gr2) → Friends of Friends
↓ ↓ ↓
"kuca-N" "velika-ADJ" "stara-ADJ"
"mala-ADJ" "nova-ADJ"
"lijepa-ADJ" "moderna-ADJ"
- Frequency: Raw count in corpus
- Relative Frequency: Per million tokens
- Grammatical Relations: Available relation types
- Node Count: Total number of nodes
- Edge Count: Total number of edges
- Density: Network connectivity measure
- Average Degree: Mean connections per node
- Degree Centrality: Number of direct connections
- Weighted Degree: Sum of edge weights
- Betweenness Centrality: Bridge importance
- Eigenvector Centrality: Influence based on neighbors
- SLI (Semi-Local Importance): Custom centrality measure
- High Degree: Well-connected nodes
- High Betweenness: Bridge nodes between communities
- High Eigenvector: Influential nodes in important neighborhoods
- Louvain: Fast community detection
- Leiden: Improved community quality
- CPM (Constant Potts Model): Resolution-based clustering
- Resolution: Controls community size (0=large, 1=small)
- Minimal Size: Minimum community size
- Quality Threshold: Community quality measure
- Community Count: Number of detected communities
- Community Sizes: Distribution of community sizes
- Modularity: Quality measure of community structure
- Coverage: Multi-lingual sentiment lexicon
- Values: Polarity scores and labels
- Integration: Original vs. network-derived values
- Coverage: English sentiment lexicon
- Values: Positive/negative scores
- Features: Continuous sentiment values
- Coverage: English synset-based lexicon
- Values: Positive/negative/objective scores
- Features: WordNet integration
- Source: Direct dictionary lookup
- Usage: Baseline sentiment scores
- Display: Tabular format with polarity information
- Source: Network-based calculation
- Method: Sentiment propagation through graph
- Parameters:
- Limit F: Number of friends to consider
- Limit FoF: Number of friends-of-friends
- Centrality Measure: For sentiment propagation
- Difference Calculation: ODV vs. ADV comparison
- Visualization: Gradient coloring for easy comparison
- Interpretation: Network influence on sentiment
- Source Lemma: Initial sentiment assignment
- Friend Network: Sentiment spread to first-degree neighbors
- FoF Network: Sentiment spread to second-degree neighbors
- Aggregation: Combined sentiment scores
- Centrality Measure: How to weight node importance
- Propagation Method: Linear vs. exponential decay
- Threshold: Minimum sentiment for propagation
- Force-Directed (kk): Spring-based layout
- Fruchterman-Reingold (fr): Energy-based layout
- Large Graph Layout (lgl): Scalable layout
- Circular: Ring-based arrangement
- Directed Layout (drl): Hierarchical arrangement
- Random: Stochastic positioning
- Tree: Hierarchical tree structure
- Network Size: Large networks benefit from lgl
- Structure: Hierarchical networks use tree/drl
- Aesthetics: Force-directed for general use
- Performance: Random for quick preview
- Size Type: degree, weighted_degree, freq
- Size Range: 0.0 to 10.0 (default: 5.0)
- Scaling: Linear or logarithmic scaling
- Label Size: 1.0 to 40.0 (default: 10.0)
- Label Type: degree or weighted_degree based
- Visibility: Show/hide labels
- Edge Size: 0.0 to 10.0 (default: 1.0)
- Edge Labels: Show/hide relationship labels
- Label Size: 0.0 to 10.0 (default: 1.0)
- Color Scheme: Automatic or custom coloring
- Zoom: Mouse wheel or zoom controls
- Pan: Click and drag to move view
- Reset: Return to default view
- Node Selection: Click to select nodes
- Edge Selection: Click to select edges
- Multiple Selection: Shift+click for multiple
- Information Panel: Details for selected elements
- Checkbox: Enable 3D network view
- Rendering: WebGL-based 3D graphics
- Performance: Hardware acceleration required
- Rotation: Mouse drag to rotate
- Zoom: Mouse wheel for depth
- Pan: Right-click and drag
- Reset: Return to default orientation
- GraphML: Standard graph format
- CSV: Node and edge tables
- JSON: Structured data export
- PNG/SVG: Static image export
- Statistics: Network metrics summary
- Community Data: Clustering results
- Sentiment Analysis: Sentiment scores and comparisons
- Input List: Upload CSV with lemma-pos pairs
- Processing: Automatic analysis of all items
- Results: Aggregated analysis report
- Export: Combined results in various formats
- Memory Usage: Monitor system resources
- Processing Time: Estimate based on network sizes
- Storage: Plan for result storage
- Function Definition: Python function with specific signature
- Integration: Add to centrality measures list
- Parameters: Configurable metric parameters
- Validation: Input validation and error handling
- Custom Centrality: Domain-specific importance measures
- Similarity Metrics: Custom similarity calculations
- Clustering Quality: Custom community quality measures
Symptoms: Error messages about Neo4j connection Solutions:
- Verify Neo4j server is running
- Check connection credentials in authSettings.py
- Ensure firewall allows database port access
- Test connection with simple query
Symptoms: Sketch Engine API failures Solutions:
- Verify API credentials are correct
- Check API rate limits and quotas
- Ensure internet connectivity
- Verify corpus access permissions
Symptoms: Application crashes or slow performance Solutions:
- Reduce network size limits
- Enable data pruning options
- Use smaller corpora for testing
- Monitor system memory usage
Symptoms: Slow response times Solutions:
- Enable caching options
- Reduce visualization complexity
- Use smaller sample sizes
- Optimize network parameters
- Connection Errors: Database or API connectivity issues
- Validation Errors: Invalid input parameters
- Processing Errors: Computation or memory issues
- Display Errors: Visualization or rendering problems
- Read Error Message: Note specific error details
- Check Parameters: Verify input values and settings
- Reduce Complexity: Simplify network or analysis parameters
- Check Logs: Review application logs for details
- Restart Application: Clear cached data and restart
- Limit Co-occurrences: Start with smaller values (5-10)
- Enable Pruning: Use frequency and score thresholds
- Selective Harvesting: Only harvest necessary data
- Batch Processing: Process multiple items efficiently
- Reduce Node Count: Limit displayed nodes
- Simplify Layouts: Use faster layout algorithms
- Disable Features: Turn off unnecessary visual elements
- Update Frequency: Reduce automatic updates
- Clear Cache: Regularly clear cached results
- Monitor Usage: Watch memory consumption
- Restart Application: Periodic restarts for long sessions
- Optimize Parameters: Balance detail vs. performance
- This Manual: Comprehensive usage guide
- API Reference: Function documentation
- Tutorials: Step-by-step examples
- Video Guides: Visual demonstrations
- GitHub Issues: Bug reports and feature requests
- Discussions: Community support and questions
- Documentation: Self-service help resources
- Email Support: Direct contact for complex issues
When reporting issues, include:
- Error Message: Exact error text
- Steps to Reproduce: Detailed reproduction steps
- System Information: OS, Python version, dependencies
- Expected vs. Actual: What you expected vs. what happened
- Screenshots: Visual evidence of the problem
This user manual covers the core functionality of ConGraCNet. For advanced topics and developer information, please refer to the Developer Guide and API Reference.