Computational Analysis of Holocaust Testimony Archives
A new computational study published on arXiv (2605.21623) systematically compares over 1,600 oral survivor testimonies from the USC Shoah Foundation and the Yale Fortunoff Video Archive. Using discourse segmentation, topic modeling, and large language models, researchers quantified structural differences in testimony formats—Shoah's interviewer-guided style versus Fortunoff's open-ended approach. Results confirm earlier qualitative distinctions but also reveal significant overlaps between the collections. The study provides a scalable framework for oral history archive comparison.
Key facts
- Study analyzes over 1,600 testimonies from USC Shoah Foundation and Yale Fortunoff Video Archive
- Uses discourse segmentation, topic modeling, and LLM-based analysis
- Quantifies structuredness through topic coherence, interviewer-survivor dynamics, and question types
- Confirms structural differences but finds significant overlaps between collections
- Published on arXiv with ID 2605.21623
Entities
Institutions
- USC Shoah Foundation
- Yale Fortunoff Video Archive
- arXiv