The Shape of Testimony: A Scalable Framework for Oral History Archive Comparison

2026-05-23 · Source: cs.AI updates on arXiv.org · Field: Science & Research — Artificial Intelligence & Machine Learning, Data Science & Analytics, Social Sciences & Behavioral Studies · Depth: Advanced, extended

Summary

The study "The Shape of Testimony: A Scalable Framework for Oral History Archive Comparison" computationally analyzes over 1,600 Holocaust survivor testimonies from the USC Shoah Foundation (1,000 interviews) and Yale Fortunoff Video Archive (668 interviews). Researchers leveraged discourse segmentation, topic modeling, and large language model (LLM) analysis to quantify "structuredness" through topic coherence, interviewer–survivor dynamics, and question types. USC interviews average 23,396 words, while Fortunoff interviews average 13,622 words, making USC interviews 1.7 times longer. Findings corroborate structural differences, with USC being more guided, especially early on, but also reveal significant overlaps and similar narrative arcs, complicating the simple "structured vs. free-form" dichotomy. The work provides a scalable, replicable framework for comparative corpus analysis.

Key takeaway

For AI Scientists and Research Scientists working with large qualitative datasets, this framework offers a robust method to computationally analyze narrative structure and dynamics. You can adapt this LLM-based pipeline to compare diverse oral history archives or other complex narrative corpora, gaining scalable insights into interactional patterns and thematic evolution. Consider integrating this approach to reassess foundational claims in your field or design new citizen-science annotation platforms.

Key insights

Computational analysis reveals nuanced "structuredness" in oral histories, challenging binary classifications.

Principles

• Oral history "structuredness" is dynamic, not static.
• Institutional design shapes narrative interaction.
• Survivor agency influences testimony evolution.

Method

A pipeline uses LLMs (ChatGPT) for Q/A topic naming and common topic extraction, alongside discourse segmentation and question type classification. It divides testimonies into 15 equal chronological segments for macro-level analysis.

In practice

• Apply LLM-based topic extraction to large corpora.
• Segment long interviews into Q/A pairs for micro-analysis.
• Classify question types to assess interviewer style.

Topics

• Holocaust Studies
• Oral History Archives
• Computational Linguistics
• Large Language Models
• Discourse Segmentation
• Topic Modeling
• Narrative Analysis

Best for: AI Scientist, Research Scientist

Related on AIssential

• Evaluating LLM-Driven Summarisation of Parliamentary Debates with Computational Argumentation — A new framework evaluates LLM summaries of parliamentary debates by preserving argumentative reasoning.
• Narrative Knowledge Weaver: Narrative-Centric Retrieval-Augmented Reasoning for Long-Form Text Understanding — Narrative Knowledge Weaver (NKW) enhances RAG by modeling evolving story worlds through structured narrative assets and multi-channel evidence assembly.
• From Vajrayana Tara to Bengali Baul: A Computational Study of Lexical Transmission Across Buddhist, Shakta, and Vaishnava Traditions in Bengal — Computational linguistics can quantitatively corroborate historical claims of lexical transmission and syncretism across religious traditions.
• On GATE, Text and Social Media Analysis, and Detecting Misinformation Online — Focused research visits can significantly advance doctoral work and foster valuable academic collaborations.
• LLMs for automatic annotation of Mandarin narrative transcripts — LLMs can automate complex discourse-level annotation in non-English speech, significantly reducing time.
• From 50K to 8.2 Million in 24 Hours: Vozinha's Algorithmic Consecration and the Multilingual Making of World Cup Visibility — Algorithmic consecration occurs when public discourse and platform metrics jointly confer symbolic value and durable visibility.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.