Elamur Research⌕
☰
home / papers / 2026 / Sentiment Analysis of German Sign Language Fairy Tales

Sentiment Analysis of German Sign Language Fairy Tales

Summary

Automatic sign language translation and synthesis systems often produce flat, emotionless output. One reason is that affect in sign languages is woven into the grammar through the face and body, and it has rarely been modeled computationally. This paper tackles three-level valence recognition (negative, neutral, positive) for German Sign Language (DGS). It uses the DGS-Fabeln-1 parallel corpus of seven fairy tales, a genre chosen because it spans a wide emotional range.

Instead of hand-annotating videos, the authors label the German text of each segment with four LLMs, take a majority vote, and assign that label to the matching video clip. From each clip they extract MediaPipe blendshapes and landmarks. These are turned into motion descriptors such as velocities, accelerations, distances, bone rotations and peak frequency, which feed an interpretable XGBoost classifier.

The model reaches a mean balanced accuracy of 0.631. Feature analysis shows that hip, elbow and shoulder movement contribute roughly as much as eyebrow and mouth activity. The released DGS-Fabeln-1-SE resource offers a lightweight, explainable baseline for emotion-aware sign language technology.

Key points

  • Data: DGS-Fabeln-1 contains 7 tales and 574 text–video segments (37–150 per tale), with clips averaging 9.6 s and about 92 minutes of video in total.
  • Text labeling:
    • Labels come from GPT5, Sonic, Mistral and GPT OSS 20B combined by majority vote, with agreement of 0.781 Krippendorff's alpha.
    • The German BERT-based classifier of Guhr et al. (2020) was rejected because it tagged over 90% of chunks as neutral.
  • Model: MediaPipe motion features feed an XGBoost classifier, which reaches 0.631 average balanced accuracy. Segments with mixed emotions are detected and excluded.
  • Context: For comparison, human annotation of the ASL dataset EmoSign achieved only 0.53 alpha, and AffectGPT reached 56.18% on 3-level sentiment there, at much higher compute cost.
  • Resources:
    • Dataset: DGS-Fabeln-1-SE on Zenodo (doi:10.5281/zenodo.18879038), including LLM labels and precomputed MediaPipe features.
    • Code: GitHub, DFKI-SignLanguage/LREC2026-DGS-Fabeln-1-SE.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Linguistics & languages

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/3cyfzw6vs9oe

@inproceedings{nunnari2026sentiment,
  title = {{Sentiment Analysis of German Sign Language Fairy Tales}},
  author = {Nunnari, Fabrizio and Jain, Siddhant and Gebhard, Patrick},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/3cyfzw6vs9oe},
  url = {https://doi.org/10.63317/3cyfzw6vs9oe}
}

References

24 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Buechel Sven and Hahn Udo. 2016. Emotion Analysis as a Regression Problem - Dimensional Models and Their Implications on Emotion Representation and Metrical Evaluation. In Frontiers in Artificial Intelligence and Applications. IOS Press. find
  2. Necati Cihan Camgoz, Simon Hadfield, Oscar Koller, Hermann Ney, and Richard Bowden. 2018. Neural Sign Language Translation. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). find
  3. Santiago Castro. 2017. Fast Krippendorff: Fast computation of Krippendorff’s alpha agreement measure. Publication Title: GitHub repository. find
  4. Tianqi Chen and Carlos Guestrin. 2016. XGBoost: A Scalable Tree Boosting System. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pages 785–794, San Francisco California USA. ACM. find
  5. Phoebe Chua, Cathy Mengying Fang, Takehiko Ohkawa, Raja Kushalnagar, Suranga Nanayakkara, and Pattie Maes. 2025. EmoSign: A Multimodal Dataset for Understanding Emotions in American Sign Language. Version Number: 1. find
  6. David M. Eberhard, Gary F. Simons, and Charles D. Fennig. 2025. Ethnologue: Languages of the World, 28 edition. SIL International. find
  7. Oliver Guhr, Anne-Kathrin Schumann, Frank Bahrmann, and Hans Joachim Böhme. 2020. Training a Broad-Coverage German Sentiment Classification Model for Dialog Systems. In Proceedings of the Twelfth Language Resources and Evaluation Conference, pages 1627–1632, Marseille, France. European Language Reso… find
  8. Annika Herrmann, Sarah Schwarzenberg, Thomas Finkbeiner, Nina-Kristin Meister, and Markus Steinbach. 2024. Expressing Emotions in Sign Languages (ExEmSiLa 2024). find
  9. Berenike Herrmann and Jana Lüdtke. 2023. A Fairy Tale Gold Standard. Annotation and Analysis of Emotions in the Children’s and Household Tales by the Brothers Grimm. Medium: HTML ,XML ,PDF Version Number: 1.0. find
  10. Jannis Klähn, Janos Borst-Graetz, and Manuel Burghardt. 2025. From dictionaries to LLMs – an evaluation of sentiment analysis techniques for German language data. Computational Humanities Research, 1:e4. find
  11. Klaus Krippendorff. 2019. Content analysis: an introduction to its methodology, fourth edition edition. SAGE, Los Angeles London New Delhi Singapore Washington DC Melbourne. find
  12. Camillo Lugaresi, Jiuqiang Tang, Hadon Nash, Chris McClanahan, Esha Uboweja, Michael Hays, Fan Zhang, Chuo-Ling Chang, Ming Guang Yong, Juhyun Lee, Wan-Teh Chang, Wei Hua, Manfred Georg, and Matthias Grundmann. 2019. MediaPipe: A Framework for Building Perception Pipelines. arXiv:1906.08172 [cs]. Ar… — [arXiv:1906.08172](https://arxiv.org/abs/1906.08172) find
  13. Albert Mehrabian. 1996. Pleasure-arousal-dominance: A general framework for describing and measuring individual differences in Temperament. Current Psychology, 14(4):261–292. find
  14. Saif Mohammad. 2011. From Once Upon a Time to Happily Ever After: Tracking Emotions in Novels and Fairy Tales. In Proceedings of the 5th ACL-HLT Workshop on Language Technology for Cultural Heritage, Social Sciences, and Humanities, pages 105–114, Portland, OR, USA. Association for Computational Lin… find
  15. Fabrizio Nunnari, Eleftherios Avramidis, and Cristina España-Bonet. 2024a. DGS-Fabeln-1. find
  16. Fabrizio Nunnari, Eleftherios Avramidis, Cristina España-Bonet, Marco González, Anna Hennes, and Patrick Gebhard. 2024b. DGS-Fabeln-1: A Multi-Angle Parallel Corpus of Fairy Tales between German Sign Language and German Text. In Proceedings of the 2024 Joint International Conference on Computational… find
  17. Osondu Oguike and Mpho Primus. 2025. Using Deep Learning Models for Multimodal Sentence Level Sentiment Analysis of Sign Language. Forum for Linguistic Studies. find
  18. F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, and E. Duchesnay. 2011. Scikit-learn: Machine Learning in Python. Journal of Machine Learning Research, 12:282… find
  19. Daniel Preoţiuc-Pietro, H. Andrew Schwartz, Gregory Park, Johannes Eichstaedt, Margaret Kern, Lyle Ungar, and Elisabeth Shulman. 2016. Modelling Valence and Arousal in Facebook posts. In Proceedings of the 7th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis,… find
  20. Viktor Suter and Miriam Meckel. 2024. Using GPT-4 for Text Analysis: Insights from English and German Language News Classification Tasks. ICWSM, US. find
  21. Şeyma Takır, Barış Bilen, and Doğukan Arslan. 2025. Sentiment analysis in sign language. Signal, Image and Video Processing, 19(3):223. find
  22. Ekaterina P. Volkova, Betty Mohler, Detmar Meurers, Dale Gerdemann, and Heinrich H. Bülthoff. 2010. Emotional Perception of Fairy Tales: Achieving Agreement in Emotion Annotation of Text. In Proceedings of the NAACL HLT 2010 Workshop on Computational Approaches to Analysis and Generation of Emotion… find
  23. Weizhe Wang and Hongwu Yang. 2021. Towards Realizing Sign Language to Emotional Speech Conversion by Deep Learning. In 2021 12th International Symposium on Chinese Spoken Language Processing (ISCSLP), pages 1–5, Hong Kong. IEEE. find
  24. Jiangtao Zhang, Qingshan Wang, and Qi Wang. 2024. U-Shaped Distribution Guided Sign Language Emotion Recognition With Semantic and Movement Features. IEEE Transactions on Affective Computing, 15(4):2180–2191. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/nunnari-2026-sentiment-analysis-german-sign-language/