FIAT/IFTA World Conference 2026

Many waves in technology development are stories of powerful entities becoming more powerful.  However, our work with audiovisual archives aims technology development along an alternative trajectory: rather than reinforcing the global monoculture, preserving the diversity of human experience and reducing the risk that less that less visible communities are absent from the historical record.


In this presentation, we describe work that follows this alternative trajectory to address such “archival absences". First, we introduce the CLAMS platform for computational processing of audiovisual archives. CLAMS is an open-source software designed to operate on modest, local hardware, permitting data sovereignty. With this computational context, we present two case studies from the collaboration between archivists at the American Archive of Public Broadcasting and researchers at Brandeis University on using technology to repair archival absences.  


The first case study focuses on cataloging items in broadcast television archives, demonstrating a multi-stage AI processing pipeline used by archivists to surface lesser-known individuals who appeared in old broadcast television programs in order to enrich metadata and improve discoverability.


The second case study focuses on linguistic absence and diversity. While ASR systems such as Whisper yield high quality transcripts of speech in high-resource languages, these systems offer no benefit for the unsupported languages and even produce misleading and nonsensical outputs. We present ongoing work on language identification for multilingual broadcast content, focusing on indigenous languages such as Yupik and Samoan. This includes challenges in segmentation, annotation, and evaluation, as well as the design of annotation workflows in collaboration with community experts.
These case studies illustrate how computational tools, when designed in collaboration with archivists and communities, can support more inclusive and representative audiovisual archives.  We close the presentation with an invitation to consider other sources of archival absence that can be resisted through creative collaborations between archivists and AI.