Field Class - Weightings

View the classification weightings associated with this Field Class.

Remarks

The Field Class Weightings tab provides a detailed, interactive workspace for reviewing the contextual features and their TF-IDF weightings used by the selected Field Class in Grooper. This tab is essential for solution designers and administrators who need to understand, validate, and troubleshoot the feature-based classification logic that determines how candidate values are scored and selected in context-aware extraction scenarios.

Purpose

Use this tab to view the list of features (such as words, phrases, or tokens) and their associated weightings that have been learned from positive and negative training examples for the current Field Class. Weightings reflect the importance of each feature in distinguishing correct values from incorrect ones, and are used by the classifier to select the best candidate during extraction.

UI Components

• Language Drop-down:
  • Displays all languages for which training data has been saved for this Field Class.
  • Select a language to view the corresponding set of features and weightings.
  • The list is automatically populated based on available training data.
• Positive Class Weightings:
  • Shows all features and their calculated weightings for the positive class (features predictive of correct values).
  • Features are sorted by importance, with the highest-weighted features at the top.
  • Columns include feature name, weighting score, count, class frequency, and other relevant metadata.
  • Supports scrolling, searching, and sorting to help users analyze the feature set.
• Negative Class Weightings:
  • Shows all features and their calculated weightings for the negative class (features predictive of incorrect values).
  • Same columns and sorting as the positive class, enabling direct comparison.
• Toolbar and Buttons:
  • {purge}: Remove all training data for the current Field Class and language.
  • Button states and tooltips provide guidance for available actions and indicate when training data is present.

Workflow and Scope

• The tab is enabled for Field Classes that have training data for one or more languages.
• All languages with available training data are listed in the drop-down; selecting a language updates both weightings lists.
• Use this tab to review which features are most influential in classification, compare positive and negative predictors, identify potential issues, or decide when to retrain or purge training data.

Example

The following diagram illustrates the typical layout of the Field Class Weightings tab, with columns matching the available feature metadata:

┌─────────────────────────────────────────────────────────────────────────────────────────┐
│ Field Class Weightings                                  {purge} │ Language: {Drop-down} │
├─────────────────────────────────────────────────────────────────────────────────────────┤
│ {PositiveClass} Weightings List                                                         │
│ ┌─────────────────────────────────────────────────────────────────────────────────────┐ │
│ │ Feature   | Weighting | Count | CWF | CTC | CF       | TF      | IDF     | ...      │ │
│ │-----------|-----------|-------|-----|-----|----------|---------|---------|----------│ │
│ │ Invoice   | 0.980000  |  120  |  3  |  80 | 0.800000 | 0.500000| 1.200000| ...      │ │
│ │ Amount    | 0.850000  |   95  |  2  |  60 | 0.600000 | 0.400000| 1.100000| ...      │ │
│ │ Due       | 0.750000  |   80  |  2  |  55 | 0.550000 | 0.350000| 1.050000| ...      │ │
│ │ ...       | ...       |  ...  | ... | ... | ...      | ...     | ...     | ...      │ │
│ └─────────────────────────────────────────────────────────────────────────────────────┘ │
│ {NegativeClass} Weightings List                                                         │
│ ┌─────────────────────────────────────────────────────────────────────────────────────┐ │
│ │ Feature   | Weighting | Count | CWF | CTC | CF       | TF      | IDF     | ...      │ │
│ │-----------|-----------|-------|-----|-----|----------|---------|---------|----------│ │
│ │ Invoice   | 0.120000  |   20  |  1  |  10 | 0.100000 | 0.050000| 0.200000| ...      │ │
│ │ Amount    | 0.250000  |   35  |  1  |  15 | 0.150000 | 0.080000| 0.300000| ...      │ │
│ │ ...       | ...       |  ...  | ... | ... | ...      | ...     | ...     | ...      │ │
│ └─────────────────────────────────────────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────────────────────────────────┘

• {Drop-down}: Select the language for which to view feature weightings.
• {PositiveClass}: Table of features and their weightings for the positive class.
• {NegativeClass}: Table of features and their weightings for the negative class.
• {purge}: Button to remove all training data for the current field class and language.

Accessibility

• Fully navigable by keyboard and mouse.
• Visual cues and highlighting indicate the selected language and feature importance.
• Tooltips and context menus provide guidance for available actions.

Advanced Features

• Dual Class Comparison: Instantly compare positive and negative predictors for the selected field class and language.
• Language Awareness: Supports multilingual training data, allowing review of features for each language independently.
• Live Updates: The weightings lists update instantly when a new language is selected or when training data changes.
• Integration: Works in conjunction with training, classification, and model management features to ensure transparency and control over the classification process.

Command Buttons

Button	Shortcut Key		Summary
delete purge		►	Purge all training. The purge button removes all training data for the current Field Class and selected language. Use this command to clear out existing training data if it is outdated, incorrect, or needs to be replaced. The button is only enabled when training data is present for the selected language. A confirmation dialog may appear to prevent accidental data loss. Tooltips provide guidance on the impact of purging training data. After purging, both the positive and negative class weightings lists will be empty until new training samples are provided and processed.

Child Controls

Name	Type		Summary
Positive Class	Weightings List	►	Displays weightings for the positive class. The Positive Class control presents all contextual features and their calculated TF-IDF weightings for the positive class (features predictive of correct values) in the selected Field Class and language. Features are sorted by importance, with the highest-weighted features at the top. Columns include: Feature, Weighting, Count, CWF (Class With Feature), CTC (Class Training Count), CF (Class Frequency), TF (Term Frequency), and IDF (Inverse Document Frequency). Supports scrolling, searching, and sorting to help users analyze the feature set. Visual cues and highlighting indicate the most influential features for classification. Tooltips and column headers provide guidance on the meaning of each value. Instantly updates when a new language is selected or training data changes. Use this control to review which features are most predictive of correct values and to validate the quality of your positive training data.
Negative Class	Weightings List	►	Displays weightings for the negative class. The Negative Class control presents all contextual features and their calculated TF-IDF weightings for the negative class (features predictive of incorrect values) in the selected Field Class and language. Features are sorted by importance, with the highest-weighted features at the top. Columns include: Feature, Weighting, Count, CWF (Class With Feature), CTC (Class Training Count), CF (Class Frequency), TF (Term Frequency), and IDF (Inverse Document Frequency). Supports scrolling, searching, and sorting to help users analyze the feature set. Visual cues and highlighting indicate the most influential features for negative prediction. Tooltips and column headers provide guidance on the meaning of each value. Instantly updates when a new language is selected or training data changes. Use this control to review which features are most predictive of incorrect values and to validate the quality of your negative training data.

Used By

Design Tab Host