Identifying Narrative Patterns and Outliers in Holocaust Testimonies Using Topic Modeling

TL;DR

This paper tackles the challenge of analyzing a large corpus of Holocaust survivor testimonies by applying contextualized topic modeling (BERTopic) to structured question–answer sections. The authors align testimonies into a fixed 15-part narrative to uncover a typical narrative schema and to compare narrative trajectories across age and gender groups. They introduce a novel outlier-detection method for divergent narratives using a score $S(t, C_A, C_B) = R_B - R_A$, with $R_B = \sum_{(i,j)∈C_B} t_i[j] · |Tvalue_B(i,j)|$ and $R_A = \sum_{(i,j)∈C_A} t_i[j] · |Tvalue_A(i,j)|$, selecting testimonies via $\arg\max_{t∈A} S(t, C_A, C_B)$. The findings illustrate a common narrative structure and meaningful differences by age and gender, demonstrating NLP’s potential to illuminate historical discourse and identify deviations in survivor narratives.

Abstract

The vast collection of Holocaust survivor testimonies presents invaluable historical insights but poses challenges for manual analysis. This paper leverages advanced Natural Language Processing (NLP) techniques to explore the USC Shoah Foundation Holocaust testimony corpus. By treating testimonies as structured question-and-answer sections, we apply topic modeling to identify key themes. We experiment with BERTopic, which leverages recent advances in language modeling technology. We align testimony sections into fixed parts, revealing the evolution of topics across the corpus of testimonies. This highlights both a common narrative schema and divergences between subgroups based on age and gender. We introduce a novel method to identify testimonies within groups that exhibit atypical topic distributions resembling those of other groups. This study offers unique insights into the complex narratives of Holocaust survivors, demonstrating the power of NLP to illuminate historical discourse and identify potential deviations in survivor experiences.

Figures (7)

• Figure 1: Testimonies number of words and number QA-pairs histogram.
• Figure 2: Corpus level QA-s topics histogram.
• Figure 3: The 5 most prevalent topics and topics variance for each part. A.T.J = Attitude toward Jews, S.P = Self-Presentation, C.L = Camps Liberation, L.P = Life Perspective, and C.M = Childhood Memories
• Figure 4: Adults vs. young survivors typical testimony t-test. The Black Point represents values with p-values under 0.01.
• Figure 5: Men vs. women survivors typical testimony t-test. The Black Point represents values with p-values under 0.01.