Nonvisual Support for Understanding and Reasoning about Data Structures

Paper Title

Nonvisual Support for Understanding and Reasoning about Data Structures

Publication Info

• Topic area: Accessibility in computer science education for blind and visually impaired (BVI) learners.
• Keywords: Accessibility, data structures, tactile graphics, screen readers, multimodal learning, binary trees, arrays, computer science education, nonvisual representation, assistive technology.

Background and Problem

• Problem / challenge: Current accessibility approaches for BVI learners in computer science education often focus on visual diagrams, which fail to preserve the structural information crucial for understanding data structures. Existing solutions are either too costly, too complex, or fail to scale effectively.
• Significance: Equitable access to computer science education is critical for fostering inclusivity and enabling BVI learners to participate fully in computing fields. Addressing these barriers can improve educational outcomes and career opportunities for BVI individuals.
• Motivation and related work: Prior research has explored tactile and auditory alternatives for visual diagrams, but these approaches often lack scalability or fail to capture the computational properties of data structures. Existing tools like tactile manipulatives and auditory navigation systems provide partial solutions but do not address the need for integrated, multimodal representations.

Solution

• Proposed approach: Arboretum, a web-based system that generates accessible representations of data structures in three synchronized formats: tabular, navigable, and tactile.
• Novelty:
  1. Automates the creation of multimodal, nonvisual representations from text-based diagram specifications.
  2. Preserves computational properties of data structures, such as roles, relationships, and traversal patterns.
  3. Integrates multiple complementary modalities to support diverse learning needs.
  4. Provides a scalable, low-barrier solution for educators to create accessible instructional materials.
• Procedure and key techniques:
  • Input: Accepts diagram specifications in languages like Mermaid and Graphviz DOT.
  • Translation: Compiles input into an intermediate representation (IR) that encodes structural and semantic relationships.
  • Output: Generates synchronized tabular, navigable, and tactile representations from the IR, ensuring consistency across modalities.

Results

• Concrete findings:
  • Participants achieved high accuracy on tasks involving arrays (91.67%) and binary trees (87.5% for leaf node identification, 100% for parent-child identification).
  • Binary search tasks were completed with 100% accuracy.
  • Tactile representations were overwhelmingly preferred, with participants rating them 5.0/5 for clarity and ease of use.
• Advantage over baselines:
  • Arboretum’s tactile graphics provided intuitive spatial understanding, outperforming traditional ALT text and screen reader-based navigation.
  • Multimodal integration allowed participants to cross-verify information, enhancing confidence and comprehension.
• Experiments / evaluation:
  • Conducted a Wizard-of-Oz study to derive design requirements for nonvisual representations.
  • Evaluated Arboretum with 8 BVI participants using tasks involving arrays and binary trees.
  • Metrics included accuracy, task completion time, and Likert-scale ratings for comprehension and confidence.
• Limitations and future work:
  • Limited to arrays and binary trees; future work should extend to advanced data structures like graphs and linked lists.
  • Screen reader support was tested only with JAWS, NVDA, and VoiceOver; broader compatibility testing is needed.
  • Tactile representations are constrained by production costs and physical limitations, making them less scalable for large or complex diagrams.

Summary

This paper introduces Arboretum, a system designed to make data structures accessible to BVI learners through synchronized tabular, navigable, and tactile representations. The system automates the generation of these formats from text-based diagram specifications, preserving structural and computational properties. Evaluation with BVI participants demonstrated high task accuracy, strong preference for tactile graphics, and the benefits of multimodal integration. Arboretum addresses key barriers in accessible CS education and provides a scalable solution for educators, with future work aimed at expanding its applicability to more advanced data structures and broader assistive technologies.