Streamlining and integrating incident data

Road incident reporting by multiple parties can lead to duplication in records and variations in descriptions, event types, locations, and timelines. This project aims to develop a robust, automated approach to incident records by leveraging advanced AI methods.

By utilising advanced natural language processing, data analysis and machine learning techniques, and uncertainty estimation, the approach will reconcile disparate information sources, resolve conflicts, extract key incident information, and produce reliable, consistent incident documentation.

The outcome of this project is expected to yield reliable and consistent incident records, providing accurate statistics and better insights for enhanced decision-making.

Participants

• Transport for NSW
• University of Technology Sydney
  

Project background

Road incident data is manually recorded and updated in various systems for state roads, motorways, and local roads by stakeholders such as the police, area operators, private motorway operators, and traffic controllers.

These records typically include key information such as incident start and end times, descriptions, event type and subtype, lane details, and location data.

The involvement of multiple parties in incident reporting often results in duplicate or multiple entries of the same incident within the systems. The issue is further complicated when entries of the same incident carry varying descriptions, event types, subtypes, locations, or start and end times.

The lack of a standardised approach to manage incident data has led to inconsistent and inaccurate information across the source systems. This duplication and fragmentation result in incorrect statistics and unreliable insights, which hinder data-driven decision-making, compromise the quality and reliability of incident data, and make effective analysis difficult for generating actionable insights.

Project objectives

Development of a mechanism that identifies and analyses potential duplicate entries of the same incident, creating a consolidated ‘golden’ record based on criteria such as incident description, type, sub-type, location, start time, and end time.

Please note …

This page will be a living record of this project. As it matures, hits milestones, etc., we’ll continue to add information, links, images, interviews and more. Watch this space!