Uncovering Patterns in Reviewers' Feedback to Scene Description Authors

Area of Science:

• Human-Computer Interaction
• Accessibility Research
• Digital Media Studies

Background:

• Audio descriptions (ADs) enhance video accessibility for blind and visually impaired users.
• Current AD generation often involves novices receiving feedback from reviewers, which is not real-time.
• The need for efficient and immediate feedback mechanisms in collaborative AD authoring systems is critical.

Purpose of the Study:

• To investigate the potential for automating reviewer feedback in the audio description creation process.
• To analyze comment patterns from novice AD authors receiving feedback from sighted or blind reviewers.
• To identify actionable insights for designing future automated feedback systems in collaborative authoring.

Main Methods:

• Analysis of 1,120 comments provided to 40 sighted novices by sighted or blind reviewers.
• Categorization of feedback into four themes: Quality, Speech Act, Required Action, and Guidance.
• Examination of feedback patterns to determine suitability for automation.

Main Results:

• Feedback provided by reviewers can be classified into four distinct themes: Quality, Speech Act, Required Action, and Guidance.
• Specific patterns within these themes show potential for automated analysis and feedback generation.
• Understanding these patterns is crucial for developing effective automated assistance tools.

Conclusions:

• Automating parts of the reviewer feedback process can streamline AD creation and improve efficiency.
• The identified themes (Quality, Speech Act, Required Action, Guidance) offer a framework for designing automated feedback systems.
• Future collaborative authoring systems can leverage these findings to provide real-time, automated support for AD creators.