"Caption it in an Accessible Way That is Also Enjoyable": Characterizing User-Driven Captioning Practices on TikTok

Title of the Paper

“Caption It in an Accessible Way That Is Also Enjoyable”: Characterizing User-Driven Captioning Practices on TikTok

Paper Information

• Research Area: Human-Computer Interaction and Video Content Accessibility
• Keywords: Video captions, TikTok, accessibility, user-generated content, open captions, social media design, disability community, multimodal data, automatic captions, standardized caption design

Research Background and Problem Statement

• Issues and Challenges: The rapid proliferation of user-generated video content has posed accessibility challenges, particularly for deaf and disabled communities. Such content is not regulated by existing media captioning laws, creating difficulties for users who rely on captions to access videos. On TikTok, while captioning culture is gaining traction, the quality and consistency of user-generated captions vary significantly.
• Research Significance: The widespread adoption of accessible captions can help deaf individuals, people with hearing impairments, neurodiverse groups, and others who rely on captions to better engage with the digital ecosystem. This is especially crucial for social interaction and entertainment consumption.
• Research Motivation and Related Work: Existing studies primarily focus on caption quality in traditional video domains (e.g., television and film), with limited attention to social media platforms like TikTok. This study examines TikTok culture to analyze how creators generate captions and evaluates their impact on users who rely on captions, addressing a gap in this area of research.

Solution

• Methodology and Research Design:
  1. Data Analysis: Content analysis of 300 TikTok videos, with 150 videos targeting general audiences and another 150 from communities related to deafness and disability (filtered using specific hashtags such as #Deaf and #Accessibility).
  2. User Interviews: Semi-structured interviews with 9 TikTok users who rely on captions due to deafness, disabilities, or related conditions.
• Innovative Contributions:
  • A novel combination of data analysis and user interviews to understand TikTok user-generated captioning practices, addressing the research gap in emergent social video captioning.
  • Proposes specific steps for standardizing captions in user-generated content, including empirical exploration of style design, textual content, and caption layout.
• Core Techniques and Steps:
  • Mixed-methods approach for data collection and analysis.
  • Development of a coding scheme to systematically analyze captions across three dimensions: audio language coverage, style and layout, and caption content.
  • Use of Word Error Rate (WER) to quantify caption accuracy.
  • Comparison with existing DCMP captioning standards to identify strengths and weaknesses of TikTok captions.

Research Findings

• Key Results:
  1. TikTok captions generally meet user needs in terms of audio language coverage and most aspects of style and layout, but improvements are needed in font size, color, and caption placement.
  2. Creators use captions for innovative expressions (e.g., using emojis to describe tone), but errors and inconsistencies in content remain.
  3. Users generally find TikTok captions adequate but believe that higher-quality, standardized captioning practices could further enhance accessibility.
• Strengths: The current open captioning culture has achieved a certain level of accessibility, but the study’s recommendations can further improve caption quality and reduce visual distractions.
• Experimental or Evaluation Results:
  • 18.3% of videos used non-verbatim captions.
  • 23.7% of videos included additional context through elements like emojis.
  • Caption error rates were relatively low but still required improvement (19.7% of videos contained at least one error).
• Limitations and Future Directions:
  • The sample size was limited to 300 videos, necessitating further expansion.
  • The study focused primarily on the U.S. English context; future research could explore captioning practices in other languages and cultural settings.
  • Recommendations include exploring how platform design can guide creators’ captioning behaviors, such as developing more user-friendly captioning tools and enhancing customization options.

Discussion and Future Outlook

• Standardization Exploration: By comparing existing U.S. standards (DCMP), the study suggests that future social video captioning standards may need to adopt more flexible coverage of music and non-verbal sounds to reduce cognitive load. It also recommends adjusting the inclusivity of non-verbatim captions based on content moderation algorithms (e.g., TikTok’s review mechanisms).
• Platform Optimization and User Engagement: Involve video creators in the design process to introduce captioning tools that balance creativity and standardization. Encourage the promotion of high-quality captions through principles of collective accessibility and mutual support.
• Research Expansion: Develop video quality evaluation metrics for analyzing open captions, enabling systematic assessment of complex captioning environments and providing guidance for video creators to optimize captions.

Conclusion

This paper provides the first comprehensive analysis of user-generated captioning practices on TikTok and reveals users’ needs and expectations for caption quality through interviews. The study offers detailed empirical insights to enhance accessibility on social platforms and proposes recommendations for standardized caption design and tool development, aiming to achieve comprehensive accessibility in user-generated content.