Computation of amendment similarity scores

Use case ID: 005

Author: IT Department, Chamber of Deputies of Italy

Date: 22 May 2024

Objective: 

Generate evidence about which amendments are similar to each other and the degree of similarity between them.

Actors: 

• Business units
• Dossier of amendments 

Prerequisites:

• Amendments stored in a structured way inside a database
• Dossier of amendments stored in a content management system

Scenario:

1. The amendments are submitted by MPs, received by offices, stored in a structured way inside a database, numbered according to presentation time and ordered according to voting order.
2. The dossier of amendments is produced and stored in a structured way (i.e. JSON format).
3. The similarity algorithm, based on semantic distance computation, is executed. For each amendment, it returns a list of amendments having a similarity score higher than a predetermined threshold. The similarity score is also shown for each item in the list.
4. The business units keep working on the dossier of amendments (legislative drafting and inadmissibility speeches), with evidence of similarities between amendments.

Expected results:

• It takes less time to manage the dossier of amendments (i.e. the writing of inadmissibility speeches).
• A situation where amendments that have similar content are managed differently is avoided.

Potential challenges:

• There could be a few false positives; human checks are therefore needed.

Data requirements:

• Structured representation of the dossier of amendments (i.e. JSON format)
• Amendment texts and metadata stored inside a database

Integrations with other systems:

• Database containing amendment texts and metadata
• Software application used for legislative drafting and for other procedural tasks on amendments

Success metrics:

• Time needed to write inadmissibility speeches
• Improved response time in terms of identifying similar amendments
• Colleague satisfaction scores