Social multi-modal event analysis via knowledge-based weighted topic model

Abstract Along with the development of mobile Internet and social network, people’s lifestyles are also changing, and many social websites (e.g. Facebook, YouTube, and WeChat) have sprung up, which leads to the emergence of a large number of multimedia data (e.g. text, image, and video) of various events. The goal of this paper is to mine event topics efficiently from massive and unordered social media data, which is beneficial to search, browse and monitor significant social events for users or governments. In order to achieve this goal, this paper proposes a novel Knowledge-Based Multi-Modal Weighted Topic Model (KBMMWTM) for social event analysis. The proposed KBMMWTM has following advantages: (1) The proposed KBMMWTM can effectively take advantage of the multi-modality of social events jointly. (2) The proposed KBMMWTM exploits word correlation in dataset as prior knowledge to improve the performance of event topic mining. We evaluate our KBMMWTM model on a real dataset and full experiments show that our model outperforms state-of-the-art models.