“Customization is Key”: Reconfigurable Textual Tokens for Accessible Data Visualizations

Title of the Paper

"Customization is Key": Reconfigurable Content Tokens for Accessible Data Visualizations

Paper Information

• Subject Area: Data Visualization and Accessible Technology Design
• Keywords: Accessible Visualization, Screen Reader, Customization, Text Description, Hierarchical Text Structure, Data Interaction, Design Goals, Domain Adaptation

Research Background and Problem

• Identified Challenges:
  • The needs of blind and low-vision (BLV) users for data visualizations vary greatly, but current solutions are mostly static and non-customizable, such as fixed language descriptions that do not adapt to users' backgrounds or tasks.
  • Generic settings in screen readers are not well-suited for the specific needs of data visualization, making it difficult for users to efficiently access key information.
• Significance:
  • Improving the accessibility and customizability of visualization solutions can significantly enhance BLV users' autonomy and engagement with data, thereby narrowing the information gap.
• Motivation and Related Work:
  • Existing designs provide only general guidance for typical users and lack personalized support for BLV users.
  • Previous research, such as the Olli tool, offers hierarchical navigation with text descriptions, laying a foundation for this study's personalized features. However, further improvements are needed to meet users' customization needs.

Solution

• Proposed Method or Solution:
  • A conceptual model based on "content tokens" is proposed, allowing users to adjust data descriptions according to their needs. The adjustments include:
    1. Presence: Users can select the range of content presented by the screen reader.
    2. Verbosity: Control the level of detail in descriptions.
    3. Ordering: Adjust the sequence of content to prioritize the most important information.
    4. Duration: Define the persistence of custom changes, enabling temporary or permanent adaptations.
  • Two implementation approaches for the model: a persistent settings menu and a quick adjustment command box.
  • The model is extended on the existing Olli tool (an open-source accessible visualization tool) to support the proposed customization model.
• Innovations:
  • Data descriptions are refined into individual "content tokens," enabling users to control visualization content with high granularity.
  • Provides task-oriented semantic settings and a responsive, customizable interaction interface.
  • Combines static text hierarchy navigation (as in the Olli tool) with dynamic personalized adjustments, balancing flexibility for both short- and long-term needs.
• Implementation Steps and Key Techniques:
  1. Use a hierarchical token mechanism to divide data into multiple information blocks, allowing users to flexibly customize through attribute settings (e.g., direction, level of detail).
  2. Provide a settings menu for persistent customization and a command box for instant adjustments.
  3. Co-design with domain experts and users, refining the design based on practical feedback from BLV users.

Research Outcomes

• Specific Results:
  • Successfully implemented the token-based customization model addressing the four core design goals (Presence, Verbosity, Ordering, Duration) on the Olli tool.
  • Provided an efficient tool for users to explore data and select information that aligns with their preferences and task requirements.
• Advantages:
  • Compared to static screen reader descriptions, the customization feature significantly improves information retrieval efficiency.
  • Enhanced task adaptability benefits users with varying levels of experience and backgrounds.
  • User evaluations indicate that the tool offers an innovative approach to making data visualizations more accessible, with participants expressing overall satisfaction with the tool's task support and user-friendliness.
• Experiment or Evaluation Results:
  • In the study, 13 BLV users evaluated the tool's interactivity, learning curve, and task performance through task tests and questionnaires.
  • The settings menu and command box demonstrated their effectiveness in supporting long-term preference adjustments and temporary task needs, respectively. User evaluations on a Likert scale ranged from 3.31 to 4, indicating general approval of the solution.
  • Quantitative data analysis revealed that users preferred concise content settings (low verbosity) but were inclined to switch flexibly for task-specific needs.
• Limitations and Future Directions:
  • The learning curve for users is steep, with some inexperienced users finding the tool complex to learn and use.
  • The current tool's explanation of data context (e.g., historical events) requires further exploration to enhance interpretability.
  • Future research may extend the model to other sensory modalities (e.g., auditory and tactile) to further improve the accessibility of multimodal data visualizations.

Conclusion

• Summary:
  • This paper proposes a hierarchical visualization customization interaction model based on content tokens, demonstrating its potential to enhance BLV users' autonomy, task adaptability, and information retrieval efficiency.
• Future Work:
  • Explore the integration of higher-level semantic construction into the tool.
  • Investigate extending the model to non-textual forms of expression, such as sonification and tactile graphics, to meet diverse user needs.