Backtracing Social Events of Interest via Logical Correlation Using GDELT

In response to the inability of mainstream methods in existing social event backtracing studies to address the logical correlation problem, the insufficient and inaccurate utilization of event characteristics for implementing logical association backtracing events, we propose a backtracing method for social events of interest. The basic idea is to obtain a set of related events given a small amount of labeled data on interested events, and to construct the causal chain of event occurrence through their merging and fusion. Specifically, we design a node feature construction strategy to select topic and title information from various elements of events to construct node features, thereby addressing the problem of insufficient and inaccurate node feature composition. We also design an adjacency relationship construction strategy, which uses the co-occurrence of participants in event records as the basis for constructing adjacency relationships, and completes the final construction of adjacency relationships by considering time elements comprehensively, thereby resolving the problem of inaccurate construction of inter-event correlation relationships. We construct a multi-layer graph convolutional network to learn the associated features and achieve node classification through a normalization function, thereby solving the problem of logical association inference between events. Experimental results demonstrate that LC-BTE is effective in backtracing analysis of events related to social events of interest. Additionally, this paper explores the performance changes of the method under semi-supervised mode and conducts experimental analysis. Finally, this paper provides the causal chain of events based on the backtracing results.

I. Introduction

Social events refer to collective activities such as sports competitions, academic events, and protest events, which are organized, involve specific participants, and occur at a certain time and place [1]. Most social events are a snapshot of social life and threaten society's normal functioning. However, some social events, such as mass protests, often have a large impact on the stability and operation of society. For example, on January 6, 2021, supporters of Donald Trump violently stormed the Capitol building to protest against the election results, leading to the interruption of the session, chaotic scenes, and even causing casualties. Therefore, it is meaningful to examine the causes and causal chain of social events of interest to prevent similar incidents in the future. In the era of big data, obtaining information from the complex and diverse internet has become an important means for people to understand society. However, the quality of information varies greatly. Therefore, developing intelligent algorithms to extract the causal chain of events from the internet effectively is a precious and necessary research work that will assist professionals in clarifying the causes of events.