2026-08-31: IMLS Grant Awarded on Identifying and Preserving Web Resources Referenced in Archived U.S. Local and National Television News

By -

In collaboration with Dr. Alexander Nwala at William & Mary and Dr. Sawood Alam at the Internet Archive, we (Himarsha R. Jayanetti, Dr. Michael L. Nelson, and Dr. Michele C. Weigle) at the Web Science and Digital Libraries (WS-DL) Research Group at Old Dominion University received an award of $750,000 in federal funds from the Institute of Museum and Library Services (IMLS) beginning in August 2026. The award supports the project, “CiteCast: Linking US local & national television news to cited web resources”. 

The project focuses on identifying and preserving web resources referenced in archived U.S. local and national television news. Television news frequently points viewers to websites, social media posts, and other online resources, but those resources can change or disappear over time. The goal of this is to help connect these references in television archives with the corresponding web resources and preserve them for future research.


Figure 1: Example TV news citing web resources. (a) A March 2013 KQED (PBS) TV news clip referring to a now defunct website (noodle.org). (b) A February 2026 FoxNews TV news clip citing (trumprx.gov). (c) A CNN TV news clip citing President Trump's April 2025 post about his administration's placement of a 125% Tariff on Chinese goods. (d) An NBC TV news clip citing President Trump's June 2025 post that the US military had bombed Iranian nuclear sites.


As a first step toward this broader goal, our recent work has focused on social media references in television news. We explored how to automatically identify social media content (such as platform logos and screenshots of posts) in archived broadcasts. We first explored a YOLOv7-based detector for social media logos and later explored multimodal large language models as another approach for identifying logos and post screenshots. This work led to a manually annotated benchmark covering 36 hours of CNN, MSNBC, and Fox News programming. The dataset includes annotations for social media logos, post screenshots, and social media mentions in closed captions, providing a foundation for evaluating automated methods for finding these references in television news archives.

We are excited to build on this work and explore how television news archives can be connected to the broader web by detecting and preserving the online resources they reference. I am really happy to see this project come to life, as this work builds on my PhD dissertation research (PhD proposal slides), and I am excited to continue contributing to this work!


- - Himarsha R. Jayanetti

Comments

Post a Comment