Classification of legislative proposals and norms

Use case ID: 069
Author: Federal Senate of Brazil
Date: 2 October 2024

Objective:

Intelligent and efficient thematic classification of legislative documents (proposals and norms) based on their textual content.

Actors:

• Federal Senate quality consultancy
• Parliamentarians who draft legislative documents
• Information system: legislative norms registration system (Sigen)
  
  Prerequisites:
• Integration with the Sigen system
• Existing knowledge base of classified legislative matters
• Trained artificial intelligence (AI) model for multi-label text classification
  
  Scenario:
• A user accesses the Sigen system.
• The user selects a legislative matter to index.
• The user requests the thematic classification from the AI system.
• The system reads the content of the document and performs the classification.
• The system returns the most relevant themes to the user, ordered by confidence score.
• The user selects the correct theme(s) from those returned by the system.

Alternate flows:

• If the system does not provide all the correct themes, the user can still choose themes manually through Sigen.

Expected results:

• Correction in the classification of documents from previous years
• Efficiency in indexing legislative matters
• Legislative proposals and norms classified with higher quality
• Reduction in review time for quality consultants

Potential challenges:

• Ensuring the AI system provides accurate and relevant responses
• Multi-label problem
• Extensive number of themes
• High degree of imbalance in the training data

Data requirements:

• Knowledge base of classified legislative matters
  
  Integrations with other systems:
• Sigen system
• Analytics and reporting tools

Success metrics:

• Average response time
• Percentage of successful hits in classification suggestions
• Reduction in the number of incorrectly classified legislative documents