Accessibility Canada Act requires federally regulated organizations doing business in certain sectors and with more than 10 employees to publish accessibility plans and follow these up in a year with a progress report. the plans are required to be updated every 3 years. This is an effort to improve accommodations for people with disabilities to increase their participation and maximize their potential.
Hundreds of organizations must publish accessibility plans, each identifying unique barriers and solutions. This diversity complicates standardized monitoring and assessment of progress within mandated accessibility areas.
This project can help stakeholders - including federal regulators, inclusive service organizations, public and private sector entities- analyze accessibility progress to enable them identify gaps, and strategically realign priorities based on findings.
We implemented RAG with OpenAI models to extract data from accessibility plans and reports. Through iterative testing, we refined prompts to efficiently gather required information and calculate metrics. The structured output was stored in a delta table and used to create a PowerBI semantic model." Challenges in developing a solution using fabric.
Due to time constraints, we manually collected and uploaded sample documents rather than implementing web crawling. Token limitations required individual OpenAI model responses, leading to manual JSON file aggregation.
We implemented a practical and simple solution to automate measurement of progress made for accessibility improvements.
We are planning to automate the ingestion of all published plans and progress reports by implementing intelligent web crawling possibly through bing and calculate metrics for all accessibility areas.
Log in or sign up for Devpost to join the conversation.