Keyword expansion techniques for mining social movement data on social media

Abstract Political and social scientists have been relying extensively on keywords such as hashtags to mine movement data from media sites, particularly Twitter. Yet, prior work demonstrates that unrepresentative keyword sets can lead flawed research conclusions. Numerous expansion methods proposed increase the comprehensiveness of keywords, but systematic evaluations these lacking. Our paper fills this gap. We evaluate five diverse techniques (or pipelines) representative movements across two distinct activity levels. results guide researchers who aim use searches data. For instance, we show word embedding-based significantly outperform other even more complex newer approaches when are in normal periods. These also less computationally intensive. More importantly, observe no single pipeline identify little than half all movement-related tweets at their peak mobilization period offline. However, coverage one is used. This true pipelines selected random.