See Google Scholar for a list of my citations.
A Clarified Typology of Core-Periphery Structure in Networks
Gallagher, R. J., Young, J-G, Foucault Welles, B. (2020). arXiv preprint 2005.10191.
To address an often overlooked gap in network science methodology, we introduce a typology which distinguishes between two types of core-periphery structure in networks: layered and hub-and-spoke. By deriving Bayesian core-periphery stochastic block models, we show empirically that there is a rich diversity of core-periphery sturcture expressed across networks.
Who Says What with Whom: Using Bi-Spectral Clustering to Organize and Analyze Social Media Protest Networks
Joseph, K., Gallagher, R. J., Foucault Welles, B. (2020). Forthcoming in Computational Communication Research.
The communities that emerge during hashtag activism are only particular instances of broader publics, but those publics are often not visible in event-based online social protest data. We propose bi-spectral clustering on user timeline data to draw out these networked publics and better contextualize how communities interact during focal hashtag events.
Reclaiming Stigmatized Narratives: The Networked Disclosure Landscape of #MeToo
Gallagher, R. J., Stowell, E., Parker, A. G., Foucault Welles, B. (2019). Proceedings of the ACM: Human-Computer Interaction (PACM: HCI), CSCW.
In October 2017, the hashtag #MeToo invited thousands of women to publicly disclose their experiences of sexual violence. We trace these disclosures and show that the more disclosures that a woman saw before disclosing herself, the more likely she was to share details with her own disclosure. Our study suggests that this networked sharing of disclosures helped reduce the stigma of sharing experiences of sexual violence during the #MeToo hashtag campaign.
Divergent Discourse Between Protests and Counter-Protests: #BlackLivesMatter and #AllLivesMatter
Gallagher, R. J., Reagan, A. J., Danforth, C. M., Dodds, P. S. (2018). PLoS ONE.
Following the extrajudicial shooting of Black teenager Michael Brown in Ferguson, MO and the rise of the Black Lives Matter movement, the counter-hashtag #AllLivesMatter also emerged. Through several computational text analyses, we show the limited topical diversity of #AllLivesMatter and how it primarily acts as a conduit to hashtags deriding #BlackLivesMatter and the necessity of acknowledging long-standing racial injustices.
Anchored Correlation Explanation: Topic Modeling with Minimal Domain Knowledge
Gallagher, R. J., Reing, K., Kale, D., Ver Steeg, G. (2017). Transactions of the Association for Computational Linguistics (TACL).
Despite the utility of topic models in answering scientific questions, they are often inflexible and do not allow substantive experts to incorporate their domain knowledge. We introduce anchored CorEx, an information-theoretic topic model that can be seeded and guided with user-selected “anchor” words. We show anchored CorEx elicits coherent, domain-relevant topics without many of the strict assumptions of traditional topic models.