Sure About That Line? Approaching Confidence-Based, Real-Time Line Assignment in Reading Gaze Data
Franziska Kaltenberger, Wei-Ling Chen, Enkeleda Thaqi, Enkelejda Kasneci
TLDR
CONF-LA offers a low-latency, confidence-based method for real-time line assignment in noisy eye-tracking data, improving accuracy for reading support.
Key contributions
- Introduces CONF-LA, a low-latency, confidence-based method for real-time eye-tracking line assignment.
- Integrates reading behavior knowledge and Gaussian line likelihoods to handle noisy data.
- Achieves fast processing (0.348 ms/fixation) and reduces online-offline accuracy gap to 1-2%.
- Significantly improves accuracy (approx. 95%) on children's data, robust to reading regressions.
Why it matters
This paper tackles the challenge of reliable real-time line assignment in noisy eye-tracking data, vital for interactive reading support. CONF-LA provides a robust, low-latency solution that handles complex reading behaviors, significantly improving accuracy. This makes real-time reading assistance more practical and effective.
Original Abstract
Remote and webcam-based eye tracking in multi-line reading suffers from various noise factors and layout ambiguity, precisely where real-time reading support needs reliable, per-fixation line assignment. Prior work largely addresses this challenge post hoc or by restricting behavior (e.g., disallowing re-reading), undermining interactive use. We propose CONF-LA (Confidence-score-based Online Fixation-to-Line Assignment), a principled, low-latency approach that integrates knowledge about reading behavior and Gaussian line likelihoods over fixations to compute a posterior-line-score and defers assignments when uncertainty is high. Evaluated on existing open-source data, CONF-LA demonstrates stable performance in post hoc analysis and closes the online-offline gap (1-2 %) with a mean per-fixation latency of 0.348 ms. Our approach exhibits particular invariance toward regressions, yielding significant improvement in ad hoc median accuracies on children data (approx. 95 %) over all tested algorithms. We encourage further research in this direction and discuss possibilities for future development.
📬 Weekly AI Paper Digest
Get the top 10 AI/ML arXiv papers from the week — summarized, scored, and delivered to your inbox every Monday.