Keynote Speakers

Abstract: Adolescents with overweight can improve not only their physical fitness but also their mental health through physical activity. However, they often face multiple barriers, such as insufficient motivation and skills, low self-esteem, and negative peer pressure. In our research, we have developed an advanced virtual reality (VR) sports system specifically designed for adolescents, named REVERIE. This system integrates deep reinforcement learning and focuses on guidance, delivering an immersive virtual sports experience. REVERIE consists of three core components: a VR sports environment, VR sports guidance, and VR sports visualization. To ensure the safety of the VR sports environment, we adopt an iterative human-in-the-loop optimization process. We have proposed a two-stage deep reinforcement learning method based on templates and feedback: in the template stage, the artificial virtual agent acquires fine-grained sports skills; in the feedback stage, it is personalized to each participant. To enhance the VR sports visualization, we have integrated a sports-oriented immersion enhancement module into REVERIE. Quantitative results show that REVERIE outperforms existing methods in terms of both safety and user experience. This research highlights the potential of VR sports as an innovative sports management approach to benefit overweight adolescents.

Abstract: Large Language Models (LLMs) are transforming sentiment analysis with more accurate, interpretable, and scalable solutions. This keynote introduces three recent advances: RVISA, which integrates rationale generation and verification to enhance implicit sentiment detection; MT-ISA, which uses automatic weight learning to manage data- and task-level uncertainties in multi-task learning; and STAR, which applies stepwise task augmentation to improve aspect sentiment quad prediction. Across benchmark datasets, these methods achieve state-of-the-art results, offering practical pathways to deploy human-aligned, context-aware, and trustworthy LLM-based sentiment analysis in real-world applications.

Abstract: In 2021，Metaverse became a hot topic around the world, and 2021 is recognized as the meta year of Metaverse. This talk will introduce the research results in the field of Metaterse enabled by intelligence techniques. Metaverse is an integrated techniques including block-chain，interaction, graphics/game，artificial intelligence，network/computing power，Internet of Things）. This talk discusses these techniques briefly, then introduce some typical applications （education metaverse，culture metaverse，military metaverse and medical metaverse）. Key techniques are explained, especially the engaged AI techniques. Finally the future research directions are discussed.