Access All Areas: Making Accessibility More Accessible
summary
Claim: converting existing PDFs to truly accessible form is feasible but imperfect. The demo uses iText 7 to rewrite PDFs, calls Azure Computer Vision to auto-generate image descriptions and Translator to set natural language, and outputs a PDF/UA compliant file. On one trillion indexed PDFs the presenter estimates under 5% accessible. The pipeline produces valid accessibility metadata but image captions sometimes mislabel content (dog -> grass), so compliance is achieved while semantic accuracy remains unreliable.
AI-written summary · about
chapters (19)
- 0:00Intro
- 1:12About Ipex
- 3:36Ipex 7 SDK
- 6:14What is accessibility
- 8:33W3C
- 14:51PDF
- 16:56The challenge
- 18:51Machines cannot read
- 19:38Example
- 24:11Code
- 25:23Accessibility Checker
- 28:51Creating Accessible Documents
- 29:54The Dark Web
- 30:50XKCD
- 31:46Proposal
- 33:16Input File
- 34:50PDF File
- 37:33Accessibility Check
- 49:49Context is everything
related talks
40mAccess All Areas: Making Accessibility More Accessible
17m15 Million Opportunities — Automated Accessibility at Scale
28mFrom Documents to UI: Designing an Inclusive Digital Experience
50mQuality Control for Accessible PDFs
29mUnderstanding Accessibility
50mPretty is Accessible
36mMaking sure your PDF/UA PDF is actually PDFUA
25mSecurity and Accessibility can be friends!