Into the Dynamics of Interpersonal Relationships from Diachronic Documents: Text-Based Temporal Social Network Construction

Chieh-Yu Lee, Christian Henriot, Hua-Yuan Hsueh, Jyi-Shane Liu, Hen-Hsen Huang

Abstract: In this work, we propose an alternative approach for constructing temporal social networks from diachronic documents by leveraging both graph and textual information. Our framework utilizes a randomized relation extraction model to extract interpersonal relationships among people from long-form documents and then enhances the extracted social network through inference on a temporal knowledge graph. The effectiveness of our approach is demonstrated through experiments on two datasets, including a politician's diary and a newspaper archive, and it has potential applications in various interdisciplinary fields such as computational politics and computational history.