This is the GitHub repository for the narrative detector on social media. I designed a machine learning-based detector of narratives in social media posts around the topic "breast cancer." The logic is straightforward based on Narrative and Persuasion Theories in communication science: detecting Who (a persona), When (past period), What (a topic or an event), and Punctuation (declarative sentence vs. interrogative sentence vs. imperative sentence). It recognizes "narratives," generates labels automatically (narrative-1, otherwise-0) in massive social media posts through "learning" experts' ground-truth in codebook and narrative/persuasion theory-based rules. In the human rater's examination, accuracy is around 92% (it's even higher evaluated by cross-validation). Please check out the documentation and code for more details and our ideas. Thanks.
The release of the complete source code and the labeled large-scale Facebook post/comment textual data set that I used is subject to the permission of my collaborators.
Honglin Bao, September 2021 baohlcs@gmail.com