Selected work

Representative Papers

Examining AI-generated historical narratives and their reception through the example of history POVs on TikTok

Jun 22, 2026

This study investigates historical inaccuracies in AI-generated first-person historical narratives on TikTok and the resulting harmful user responses, with a particular focus on the proliferation of hate speech and misinformation surrounding sensitive topics such as the Holocaust. Employing a two-phase empirical approach—combining an exploratory pilot study with large-scale data analysis via the TikTok Research API—the research integrates manual annotation and a DistilBERT-based comment classification model for mixed-methods analysis. The findings reveal, for the first time in a systematic manner, that contemporary historical themes dominate and frequently contain factual errors; moreover, content related to the Holocaust elicits significantly more hate speech and misinformation compared to other topics like the Black Death, underscoring the risks posed by AI-generated historical content on short-video platforms and the urgent need for effective moderation and oversight.

0 citationsRead paper

FLEURS-Kobani: Extending the FLEURS Dataset for Northern Kurdish

Mar 31, 2026

This study addresses the scarcity of publicly available speech benchmarks for Northern Kurdish (Kurmanji, KMR), which has hindered progress in automatic speech recognition (ASR) and speech translation. To bridge this gap, the authors introduce FLEURS-Kobani, the first KMR-focused dataset derived from the FLEURS benchmark, comprising 5,162 validated utterances (18 hours and 24 minutes) from 31 native speakers, enabling research in ASR, end-to-end speech-to-text translation (S2TT), and speech-to-speech translation (S2ST). Leveraging the Whisper v3-large architecture with Common Voice pretraining and a two-stage fine-tuning strategy, the proposed system achieves a word error rate (WER) of 28.11% and character error rate (CER) of 9.84% on ASR, and a BLEU score of 8.68 for end-to-end KMR→EN S2TT, establishing a foundational benchmark for low-resource Kurdish speech processing.

0 citationsRead paper

From Consensus to Split Decisions: ABC-Stratified Sentiment in Holocaust Oral Histories

Mar 30, 2026

This study addresses the poor performance of existing sentiment polarity models on structurally complex and highly heterogeneous large-scale oral history texts—such as Holocaust testimonies—under domain shift. The authors propose an ABC hierarchical framework grounded in inter-model consistency, integrating multi-model label triangulation with emotion distribution–based auxiliary analysis. They systematically evaluate 107,305 utterances and 579,013 sentences using three pretrained Transformer-based polarity classifiers alongside a T5 emotion classifier, quantifying model disagreement through agreement rates, Cohen’s Kappa coefficients, and normalized confusion matrices. Results reveal only low-to-moderate overall model consistency, with disagreements predominantly concentrated near the neutral boundary. Crucially, emotion distributions vary significantly across ABC hierarchy levels, offering an actionable diagnostic pathway for assessing uncertainty in affective analysis of sensitive historical narratives.

0 citationsRead paper

Recent publications

Latest Papers

Examining AI-generated historical narratives and their reception through the example of history POVs on TikTok

Jun 22, 2026

This study investigates historical inaccuracies in AI-generated first-person historical narratives on TikTok and the resulting harmful user responses, with a particular focus on the proliferation of hate speech and misinformation surrounding sensitive topics such as the Holocaust. Employing a two-phase empirical approach—combining an exploratory pilot study with large-scale data analysis via the TikTok Research API—the research integrates manual annotation and a DistilBERT-based comment classification model for mixed-methods analysis. The findings reveal, for the first time in a systematic manner, that contemporary historical themes dominate and frequently contain factual errors; moreover, content related to the Holocaust elicits significantly more hate speech and misinformation compared to other topics like the Black Death, underscoring the risks posed by AI-generated historical content on short-video platforms and the urgent need for effective moderation and oversight.

0 citationsRead paper

FLEURS-Kobani: Extending the FLEURS Dataset for Northern Kurdish

Mar 31, 2026

This study addresses the scarcity of publicly available speech benchmarks for Northern Kurdish (Kurmanji, KMR), which has hindered progress in automatic speech recognition (ASR) and speech translation. To bridge this gap, the authors introduce FLEURS-Kobani, the first KMR-focused dataset derived from the FLEURS benchmark, comprising 5,162 validated utterances (18 hours and 24 minutes) from 31 native speakers, enabling research in ASR, end-to-end speech-to-text translation (S2TT), and speech-to-speech translation (S2ST). Leveraging the Whisper v3-large architecture with Common Voice pretraining and a two-stage fine-tuning strategy, the proposed system achieves a word error rate (WER) of 28.11% and character error rate (CER) of 9.84% on ASR, and a BLEU score of 8.68 for end-to-end KMR→EN S2TT, establishing a foundational benchmark for low-resource Kurdish speech processing.

0 citationsRead paper

From Consensus to Split Decisions: ABC-Stratified Sentiment in Holocaust Oral Histories

Mar 30, 2026

This study addresses the poor performance of existing sentiment polarity models on structurally complex and highly heterogeneous large-scale oral history texts—such as Holocaust testimonies—under domain shift. The authors propose an ABC hierarchical framework grounded in inter-model consistency, integrating multi-model label triangulation with emotion distribution–based auxiliary analysis. They systematically evaluate 107,305 utterances and 579,013 sentences using three pretrained Transformer-based polarity classifiers alongside a T5 emotion classifier, quantifying model disagreement through agreement rates, Cohen’s Kappa coefficients, and normalized confusion matrices. Results reveal only low-to-moderate overall model consistency, with disagreements predominantly concentrated near the neutral boundary. Crucially, emotion distributions vary significantly across ABC hierarchy levels, offering an actionable diagnostic pathway for assessing uncertainty in affective analysis of sensitive historical narratives.

0 citationsRead paper