Date of Award
Department or Program
This thesis describes a variety of approaches to developing a computational model of narrative on social media. Our goal is to use such a narrative model to identify efforts to manipulate public opinion on social media platforms like Twitter. We present a model in which narratives in a collection of tweets are represented as a graph. Elements from each tweet that are relevant to potential narratives are made into nodes in the graph; for this thesis, we populate graph nodes with tweets’ authors, hashtags, named entities (people, locations, organizations, etc.,), and moral foundations (central moral values framing the discussion). Two nodes are connected with an edge if the narrative elements they represent appear together in one or more tweets, with the edge weight corresponding to the number of tweets in which these elements coincide. We then explore multiple possible deep learning and graph analysis methods for identifying narratives in a collection of tweets, including clustering of language embeddings, topic modeling, community detection and random walks on our narrative graph, training a graph neural network to identify narratives in the graph, and training a graph embedding model to generate vector embeddings of graph nodes. While much work still remains to be done in this area, several of our techniques, especially the generation and clustering of graph embeddings, were able to identify groups of related and connected nodes that might form the beginnings of narratives. Further study of these or other techniques could allow for the reliable identification of full narratives and information operations on social media.
Bailey, Anne, "Towards a Computational Model of Narrative on Social Media" (2022). Dartmouth College Undergraduate Theses. 264.