COR Themes for Readability from Iterative Feedback

Document Title

COR Themes for Readability from Iterative Feedback

Document Information

• Subject Area: Human-Computer Interaction and Readability Optimization
• Keywords: readability, text formatting, customization, reading themes, crowdsourcing, machine learning, user experience, typography design, accessibility

Research Background and Problem

• Problem/Challenge:

  • Background: Digital reading is a primary means of information acquisition, and personalized adjustments to text formatting, such as font and spacing, can enhance the reading experience.
  • Challenges:
    1. Users often struggle to find suitable text settings on their own.
    2. Interactions between adjustment parameters are complex, e.g., character spacing may affect line spacing.
    3. Narrowly targeted customizations (e.g., fonts designed specifically for dyslexia) fail to address diverse user needs.

• Significance: Providing preset text themes suitable for most users can improve reading efficiency and comfort, while advancing the development of accessible digital reading technologies.

• Research Motivation: The study aims to develop standardized "reading theme" presets by integrating user feedback, automated algorithms, and designer expertise, addressing the complexity of user customization while promoting inclusivity in design.

Solution

• Method or Solution:

  1. Development of Reading Themes:
     • Combine text parameters such as font selection, character spacing, word spacing, and line spacing into multiple preset themes: Compact, Open, Relaxed.
  2. Feedback Loop:
     • Adopt a user-centered design approach by crowdsourcing user preference data.
     • Utilize machine learning algorithms for automatic clustering to generate reading themes.
     • Involve designers to refine the automatically generated themes.

• Innovations:

  1. Three-Stage Feedback Method: Integrates user adjustments (Stage 1), automated clustering (Stage 2), and expert refinement (Stage 3).
  2. Multi-Round Iteration: Conducted four design iterations to continuously optimize reading themes to meet diverse needs.
  3. Dynamic Theme Generation: Experimentally derived preset themes that automatically adapt to different user requirements.

• Implementation Steps and Key Technologies:

  1. User Data Collection:
     • Use a complete interactive interface to allow users to adjust text settings and collect preference data post-adjustment.
  2. Automated Clustering:
     • Apply convolutional neural networks (CNN) and k-means algorithms to cluster user-generated text formats.
     • Extract representative settings from each cluster as new theme candidates.
  3. Designer Calibration:
     • Engage multiple designers to review the automatically generated themes and supplement missing design considerations as needed.
  4. Experiment and Evaluation:
     • Evaluate the performance of the three final themes (Compact, Open, Relaxed) by comparing metrics such as reading speed, comfort, and comprehension.

Research Outcomes

• Specific Outcomes:

  1. Proposed three reading themes (Compact, Open, Relaxed) combining different font and spacing settings to cater to diverse user needs.
  2. Standardized the process of integrating user preferences and inclusive design, supported by empirical data and design consensus.

• Advantages:

  1. User Satisfaction:
     • Default theme settings reduced the time and effort required for users to adjust parameters.
     • The diversity of theme options enhanced representation of user preferences.
  2. Performance Optimization:
     • Compared to default text settings, the themes improved users' reading speed, comfort, and comprehension.
  3. Inclusivity Validation:
     • The study included users across different age groups and those with reading disabilities (e.g., dyslexia), demonstrating universality.
     • The "Relaxed" theme with larger line spacing was preferred by older users and those with reading disabilities.

• Experimental or Evaluation Results:

  1. Users achieved optimal reading speed with the Compact and Open themes, while the Relaxed theme provided the highest comfort and comprehension.
  2. Older users and those with reading disabilities significantly preferred the larger line spacing of the Relaxed theme, while younger professional users favored the compact layout of the Compact theme.
  3. After four iterations, the settings for reading themes stabilized, and the time required for user adjustments decreased progressively.
  4. Mixed-effects linear models (LME) validated the impact of themes on speed and comfort across variables such as gender, age, and learning disabilities.

• Limitations and Future Directions:

  1. Population Limitations:
     • The study primarily focused on users aged 18 and above who are native English speakers, excluding children, non-native speakers, and other languages.
  2. Scenario Constraints:
     • The research was limited to desktop digital reading scenarios and has not been extended to mobile devices or physical materials.
  3. Technical Improvements:
     • Clustering time depends on user volume and design iterations; future work should explore more efficient computational methods.
     • The font normalization process still has room for improvement, such as accounting for differences in character width.

Through this elegant and data-driven approach, the study provides a new perspective on user-centered digital reading technologies, while demonstrating the efficiency and applicability of interdisciplinary team collaboration in complex product development.