Year: 2025
Type: article
Abstract: Online storytelling is an essential channel for users to express their experiences and opinions and therefore influence online society. Yet, despite its importance, approaches to story understanding on social media have not advanced sufficiently. Specifically, current approaches can carry out high-level, aggregate analyses on a corpus of stories but do not provide a way of understanding individual stories. We consider a major source of social behavior, app reviews, which surprisingly are rarely studied in social media research. We observe that app reviews often contain one or more stories. These stories exhibit complex structures and are often presented via events that are not placed in their natural order. Accordingly, we introduce Saga, an approach that carries out a deep analysis of the event-based structures and substructures arising in app reviews. Saga’s main contribution is how it goes beyond the state of the art in identifying fine-grained story (sub)structures. In addition, it supports querying stories (and their containing app reviews) according to these (sub)structures. These specific (sub)structures help identify stories that serve different information goals. Saga is evaluated both computationally on a publicly available data source and via a human study validating the helpfulness in addressing various information goals.
Source: ACM Transactions on the Web
DOI: 10.1145/3728313
Authors: Hui Guo, Munindar P. Singh
Institution: North Carolina State University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4409483211