Standardizing Extended Descriptions: User Stories, Testing Results, and Current Limitations

This document outlines use cases for extended descriptions in EPUB and web content, addressing the need for standardized semantic markup and consistent reading system behavior. Extended descriptions provide detailed textual alternatives for complex visual content including images, diagrams, tables, videos, and mathematical formulas. The standardization of extended descriptions is important for interoperability, and user experience consistency across the digital publishing ecosystem. To validate these requirements, we conducted practical tests using different implementation approaches, with results and findings detailed in the Test Structure and Implementation section.

State of the Art

Web standards provide multiple mechanisms for extended descriptions (summary/details elements, links, aria-details attribute), but each presents limitations in publishing contexts (both in EPUB and on the web). The summary/details pattern becomes problematic in paginated content, as these elements can break across page boundaries, disrupting the user experience.

More significantly, content creators often place extended descriptions in separate HTML files to maintain content organization and reduce file size. This practice renders the aria-details attribute ineffective, since it can only reference elements within the same document. Publishers typically establish the relationship between images and external extended descriptions through standard links, but these links lack specific semantics that would allow assistive technologies or user agents to programmatically identify their purpose as extended description access points.

Similarly, extended description containers are commonly marked up as generic section elements without distinctive semantics. This absence of specific semantics makes it impossible for automated tools, quality assurance systems, or adaptive user interfaces to distinguish extended description content from other sections within the document structure.

User Stories

The following user stories demonstrate requirements that extend beyond EPUB to encompass the broader web ecosystem. While digital publishing formats like EPUB face unique challenges such as pagination and offline reading environments, the fundamental accessibility needs for extended descriptions are universal across all digital content platforms. Web applications, educational platforms, news websites, and online documentation systems all serve users with diverse abilities who require detailed descriptions of complex visual content. The semantic markup patterns and programmatic identification methods needed for extended descriptions must therefore work consistently across web standards and publishing formats to ensure a seamless accessible experience. Publishers often repurpose content across multiple channels - from EPUB to web to mobile applications - making cross-platform compatibility essential. Additionally, many modern reading systems and content management platforms are web-based, meaning that solutions developed for web accessibility directly impact the publishing workflow and user experience in digital books.

Screen Reader User

• As a screen reader user I want extended descriptions to be programmatically identifiable, so that my assistive technology can announce their availability
• As a screen reader user I want to return to the exact location where I accessed an extended description, so that I maintain my reading flow and don't lose context
• As a screen reader user I want link to extended description and the container of the extended description to be programmatically identifiable, so that my assistive technology can announce richer experience

Visual Reader

• As a dyslexic visual reader I want to easily identify when images have extended descriptions available, so that I can access additional context and information when needed
• As a visual reader I want to access extended descriptions without losing my place in the content, so that I can return to my reading position efficiently

User Agent Developer

User Agents in the publishing world are EPUB Reading Systems that are usually built on web views, or developed as web applications or are browsers plugins which enhances the browser’ functionality for reading publications.

• As a user agent developer I want standardized semantic roles for extended description markup, so that I can implement consistent user interface behaviors
• As a user agent developer I want programmatically identifiable extended descriptions, so that I can develop innovative presentation methods (sidebars, overlays, detail views)
• As a user agent developer I want programmatically identifiable extended descriptions, so I can allow readers to escape and skip them

Content Creator

• As a content creator I want clear, standardized markup patterns for extended descriptions, so that I can ensure consistency and accessibility across my content
• As a content creator I want extended descriptions to work consistently across different user agents, so that all users receive the same accessible experience
• As a content creator I want extended description patterns that work in both EPUB and general web contexts, so that I can maintain consistency across different publication formats
• As a content creator I want to programmatically identify which images have extended descriptions, so that edit my content with specific tools and UI for managing extended descriptions

Quality Assurance Tool

• As a quality assurance tool (like ACE by DAISY) I want to systematically identify extended descriptions, so that I can provide accurate extended description reports
• As a content management system I want to generate inventories of all extended descriptions in a collection of content, so that content creators can track and maintain their accessible content

Testing Results and Current Limitations

We conducted practical tests to evaluate how current web standards could address the Extended Description User Stories requirements. These tests revealed significant limitations and gaps in existing approaches.

Ecco il testo riscritto per chiarire meglio la struttura del test:

Test Structure and Implementation

We designed a comprehensive test case consisting of two interconnected documents to evaluate extended description implementation patterns.

Test Document 1: Main Content File

The primary test file simulates a typical chapter containing two images with different markup patterns:

• Image A: Enclosed within a figure element with proper figcaption, demonstrating standard semantic structure
• Image B: Standalone image without container elements, representing common real-world scenarios

Both images utilize the aria-details attribute pointing to links that navigate to their respective extended descriptions. These links are positioned immediately after each image and serve as the programmatic bridge to the extended content.

Test Document 2: Extended Descriptions File

The second document contains the actual extended descriptions.

Each extended description section follows a consistent pattern:

• The original image reproduced with role="presentation" and empty alt="" attribute as visual reference
• Detailed structured description of the visual content
• Backlink with role="doc-backlink" enabling return navigation to the source location

JavaScript Validation Layer

We implemented JavaScript testing code that validates the programmatic identification approach using the XPath selector //a[@id=//img[@aria-details]/@aria-details] to identify links pointing to extended descriptions.

Since the extended description is located on a separate page, it is not possible to use JavaScript to identify whether a given section contains an extended description or not.

Identified Limitations

Testing revealed two constraints:

1. Cross-Document Chain Breakage: The XPath-based identification system functions only within single documents. When extended descriptions reside in separate files, the programmatic relationship chain breaks, leaving containers semantically unidentifiable.

2. Computational Overhead: The XPath approach requires user agents to perform complex document preprocessing and DOM traversal operations for each extended description discovery. This creates significant computational burden, particularly in documents with multiple images or when processing large content collections.

3. Implementation Complexity: XPath-based solutions demand sophisticated parsing logic from user agents, making implementation more error-prone and resource-intensive compared to direct, imperative markup approaches.