Visually Connecting Historical Figures Through Event Knowledge Graphs



Shahid Latif, Shivam Agarwal, Simon Gottschalk, Carina Chrosch, Yanick Christian Tchenko, Felix Feit, Johannes Jahn, Tobias Braun, Elena Demidova, Fabian Beck

 External link (DOI) 

 View presentation:2021-10-28T18:00:00ZGMT-0600Change your timezone on the schedule page
2021-10-28T18:00:00Z

Exemplar figure, described by caption below — VisKonnect offers a natural language interface to search the intertwined lives of historical figures using a mix of text and visualizations. (1) It answers a query about the marriage of Marie Curie and Pierre Curie while the event timeline allows for verification of the answer as well as further exploration of their lives. (2) The answer of the query about Albert Einstein and Erwin Schroedinger lists down Solvay conference as a shared event where two scientists met. The relationship graph on the right also leads to the same information disclosing a photograph where the two scientists can be seen together.

Keywords

Domain Agnostic, General Public, Graph/Network and Tree Data, Temporal Data, Text/Document Data

Abstract

To research lives of historical figures and their interactions with other famous people of the same era, users have to usually go through long text documents. Knowledge graphs store information about entities such as historical figures and their relationships---indirectly through shared events---in a structured manner. We develop a visualization system, VisKonnect, for analyzing the intertwined lives of historical figures based on the events they participated. The users' query is parsed for identifying named entities and related data is queried from an event knowledge graph. While a short textual answer to the users' query is generated using the GPT-3 language model, various linked visualizations provide context, display additional information related to the query, and allow in-depth exploration.