Prof. Huamin Qu and CSE PhD Alumni and Students Won Four Best Paper Honorable Mention Awards in IEEE VIS 2021
In IEEE VIS 2021, Prof. Huamin QU, Chair Professor of Department of Computer Science & Engineering and also Director of Interdisciplinary Programs Office, with his students and researchers, received four Best Paper Honorable Mention Awards for his papers including "Augmenting Sports Videos with VisCommentator", "KG4Vis: A Knowledge Graph-Based Approach for Visualization Recommendation", "M2Lens: Visualizing and Explaining Multimodal Models for Sentiment Analysis" and "VBridge: Connecting the Dots Between Features and Data to Explain Healthcare Models".
The award-winning paper "Augmenting Sports Videos with VisCommentator" co-authored with CSE alumni Zhutian CHEN, Yingcai WU, researchers Shuainan YE, Xiangtong CHU, Haijun XIA and Hui ZHANG, proposes to utilize machine learning-based data extractors and a meticulously outlined design space to generate visualization recommendations, which would help sports analysts to create augmented videos. The system was tested by 7 experts in related fields who had ended up with high satisfaction after being able to create augmented sports videos within a short period of time. They believe the system will provide insightful suggestions to future improvements of sports analytics.
Means to simplify the creation of data visualization are addressed in "KG4Vis: A Knowledge Graph-Based Approach for Visualization Recommendation" co-authored with Dr. Yangqiu SONG, the Assistant Professor of Department of Computer Science and Engineering at the Hong Kong University of Science and Technology, CSE alumni Yong WANG, CSE PhD students Haotian LI and researcher Songheng ZHANG. The paper suggests the use of KG4Vis, a knowledge graph-based approach to visualize recommendation, to widen the accessibility of data visualization to general users. A TransE-based embedding technique is utilized to construct a desirable visualization order, such that powerful visualizations could be deduced for any new data set. The approach has been shown effective through multiple evaluation methods such as quantitative comparisons, case studies and expert interviews.
The investigation of how the users can achieve deeper insights from multimodal models for sentiment analysis is described in the award-winning paper "M2Lens: Visualizing and Explaining Multimodal Models for Sentiment Analysis" co-authored with CSE alumni Yong WANG, CSE PhD students Xingbo WANG, Jianben HE, Zhihua JIN and researcher Muqiao YANG. The paper indicates that the interactive visual analytics system, M2Lens , can be used to visualize and explain multimodal models for sentiment analysis. Through providing explanations on intra- and inter-modal, summarizing influence of typical interaction types on model predictions, identifying frequent, influential multimodal features and allowing multi-faceted exploration of model behaviors, it elucidate how sentiment predictions are generated from multimodal information.
Issues with model transparency and interpretability were solved with the introduction of "VBridge" in "VBridge: Connecting the Dots Between Features and Data to Explain Healthcare Models" co-authored with CSE PhD alumni Dongyu LIU, current students Furui CHENG, Yanna LIN and researchers Fan DU, Alexandra ZYTEK, Haomin LI and Kalyan VEERAMACHANENI. It is a visual analytics tool that seamlessly incorporates ML explanations into clinicians' decision-making workflow. A novel hierarchical display of contribution-based feature explanations and enriched interactions that connect the dots between ML features, explanations and data helped clinicians to better interpret and use model predictions for clinician decisions making.
IEEE Visualization Conference (VIS) is the premier international conference of data visualization and visual analytics. The conference convenes an international community of researchers and practitioners from universities, government, and industry to exchange recent findings on the design and use of visualization tools.
Congratulations to Prof. Qu, Dr. Song and CSE alumni, current PhD students and researchers!
For more details, please refer to the IEEE VIS 2021 Best Paper Awards website.