Stable Cluster Core Detection in Correlated Hashtag Graph

TL;DR

This paper proposes three algorithms to detect stable core clusters within correlated hashtag graphs on Twitter, aiding in understanding persistent versus transient hashtag groups for better event and topic tracking.

Contribution

It introduces novel algorithms for identifying stable core clusters in hashtag graphs, enhancing the analysis of evolving social media topics.

Findings

Three algorithms successfully detect stable core clusters

Algorithms outperform baseline methods in stability detection

Enhanced understanding of hashtag group dynamics

Abstract

Hashtags in twitter are used to track events, topics and activities. Correlated hashtag graph represents contextual relationships among these hashtags. Maximum clusters in the correlated hashtag graph can be contextually meaningful hashtag groups. In order to track the changes of the clusters and understand these hashtag groups, the hashtags in a cluster are categorized into two types: stable core and temporary members which are subject to change. Some initial studies are done in this project and 3 algorithms are designed, implemented and experimented to test them.