Automatically producing IMS AccessForAll Metadata
Matteo Boni, Sara Cenni, Silvia Mirri, Ludovico Antonio Muratori, Paola Salomoni · 2006 · Proceedings of the 2006 International Cross-Disciplinary Workshop on Web Accessibility (W4A) · doi:10.1145/1133219.1133237
Summary
This paper from the University of Bologna presents the ACCMDBuilder, a software component that automatically generates IMS AccessForAll Metadata (ACCMD) descriptions for accessible learning objects (LOs). The IMS AccessForAll specification provides a structured way to describe the accessibility features of e-learning content — identifying what sensory modalities a resource uses (visual, auditory, textual, tactile) and linking primary resources to their equivalent alternatives. While powerful, manually creating these metadata descriptions is complex and time-consuming, reinforcing the perception that producing accessible learning materials is too burdensome. The ACCMDBuilder was integrated into an existing authoring tool called ISA-LOB, used in the University of Bologna's A3 Project, which produced approximately 40 accessible learning objects. The paper demonstrates the system through two case studies: Flash animations (requiring alternative image sequences and text descriptions) and SMIL-based video lectures (requiring synchronized captions, textual slide descriptions, and alternative HTML versions). For video lectures, the system handles five information flows: the video track, audio track, slide sequence, caption sequence, and slide content descriptions. The authoring tools force authors to provide required alternatives (such as image descriptions) during content creation, then the ACCMDBuilder automatically structures these into compliant ACCMD XML metadata and packages everything into a SCORM-compliant learning object.
Key findings
The system successfully automated the generation of IMS AccessForAll Metadata, producing over 1,000 lines of ACCMD code for a test set containing 32 images, 4 animations, and a video lecture — demonstrating that manual metadata creation at this scale would be impractical. User testing with people with disabilities confirmed that SMIL video lectures needed multiple alternative versions: an HTML hypertext linking textual slide content to audio explanations (for blind users who experience cognitive overload from simultaneous audio tracks), and a version combining slide text with captions (for deaf users who find synchronized changing text too fast). The authors identified a gap in the IMS ACCMD standard: it lacks provisions for epileptic users who may need alternatives to content with flashing or rapid visual changes. The produced learning objects were tested across multiple browsers and assistive technologies (including IBM Home Page Reader and Lynx) and met WCAG 1.0 AA and Section 508 requirements, as well as Italian accessibility law constraints. A key practical limitation was that existing Learning Content Management Systems largely ignored the ACCMD descriptions, meaning the metadata was generated but not yet consumed for adaptive content delivery.
Relevance
This paper addresses a persistent challenge in digital accessibility: the gap between the availability of accessibility standards and their practical adoption. The insight that accessibility metadata must be generated automatically — because manual creation is too burdensome for content authors — remains highly relevant today. Modern authoring tools still struggle with this balance between thoroughness and usability. The paper's approach of forcing authors to provide alternatives at the point of content creation (rather than as a separate remediation step) reflects what is now considered best practice in accessible content workflows. The finding that blind users experienced cognitive overload from synchronized audio streams in SMIL presentations highlights the importance of user testing beyond automated compliance checks. For practitioners working on e-learning accessibility, the paper's framework for thinking about primary resources, equivalent alternatives, and supplementary content provides a useful model, even though the specific IMS specifications have evolved into the ISO/IEC 24751 AccessForAll framework.
Tags: e-learning accessibility · metadata · learning objects · multimedia accessibility · authoring tools · SCORM · content alternatives
Standards referenced: IMS AccessForAll Metadata · IMS ACCLIP · SCORM · WCAG 1.0 · Section 508 · SMIL