Real-time multi-modal social event detection: a new dataset and a key instance-driven, quality-aware graph neural network

<p dir="ltr">Social event detection (SED) involves identifying and analysing significant real-world events using data generated on social media platforms. With the rise of platforms like Weibo and Twitter, users are sharing not just text but also images, videos, and other multimedia content. However, most existing SED methods remain text-focused, limiting their ability to fully capture the complexity of real-world social dynamics. As social media content becomes more multi-modal, the need for SED methods that can integrate and analyse multiple data types has become increasingly pressing. Moreover, the lack of multimodal datasets specifically designed for SED has blocked the development of models that can effectively exploit these rich content types. To address these challenges, this thesis introduces WEIBO2022, an extensive multi-modal social event detection dataset that includes both text and image data. The dataset is available in two versions: WEIBO2022- Medium, comprising 25,435 entries and WEIBO2022-Large, containing 79,825 entries. In addition, I present a novel network called the Key Instance-driven Quality-aware Graph Neural Network (KQGNN), which features a key instance-driven library, a quality-aware learning process, and a multi-modal fusion module, enhancing its ability to detect events accurately in both offline and real-time settings. Extensive experiments demonstrate the superiority of the proposed model.</p>