Analysis of the SalChartQA Dataset

Description: SalChartQA is a human visual attention dataset with 3, 000 visualisations and 6, 000 saliency maps, each with a question, and bounding box annotations of visual elements. We plan to analyse this dataset, and find more insights between gaze behaviour and some metrics, such as question correctness and answer uncertainty.
For example, the figure contains two question-driven saliency maps. Question of the left: Which reason has the highest amount of votes which is 67? Question of the right: What is the average of all reasons?

Research Questions: 1. Can gaze data represent the correctness / uncertainty of answers (how many participants answered this question correctly, how many unique answers are there)?
2. How do viewing patterns differ under different questions and different visualisation types, especially regarding areas of interest? Does the viewing patterns change over time (e.g. the first 3 seconds vs. the last 3 seconds)?
3. Your own ideas!

Supervisor:Yao Wang and Susanne Hindennach

Distribution:20% Literature, 60% Implementation, 20% Analysis and Evaluation

Requirements:Experience with Python, ideally some knowledge of statistics.

Literature: Yao Wang, Weitian Wang, Abdullah Abdelhafez, Mayar Elfares, Zhiming Hu, Mihai Bâce, Andreas Bulling. 2024. SalChartQA: Question-driven Saliency on Information Visualisations. In Proc. ACM SIGCHI Conference on Human Factors in Computing Systems (CHI), pp. 1–14, 2024. Paper link