Interactive tool for evaluation

[omri374]

User Story

As a data scientist or analyst
I want to run an end-to-end interactive evaluation process for Presidio
So that I could quickly set up, configure, and assess Presidio's performance with synthetic data and structured evaluation

Acceptance Criteria

• Provide an interactive pipeline (e.g., Gradio or Streamlit app) that includes:
  1. Data Generation:
     • Users can define template sentences and Faker providers to generate synthetic PII data.
     • The generated dataset is structured and stored for evaluation.
  2. Presidio Configuration:
     • Users can specify configuration settings via a YAML file (aligned with AnalyzerEngineProvider).
  3. Entity Mapping:
     • Users can define a mapping between dataset entities and Presidio entities for evaluation.
  4. Presidio Evaluation:
     • The Evaluator object runs Presidio on each sample and compares predictions to ground truth.
  5. Metrics & Plots:
     • Compute and display key evaluation metrics (per-entity precision, recall, F2-score, and overall PII detection scores).
     • Generate visualizations (e.g., confusion matrices, precision-recall curves).
  6. Error Analysis:
     • Highlight common failure cases (e.g., false positives, false negatives).
     • Provide interactive filtering options to inspect misclassified samples.
• Ensure the workflow is modular so users can modify each step independently.
• Provide a sample dataset and default configurations to help users get started quickly.
• Document the process with clear instructions and examples.