Elamur Research⌕
☰
home / papers / 2026 / Is Sentiment Banana-Shaped? Exploring the Geometry and Portability of Sentiment Concept Vectors

Is Sentiment Banana-Shaped? Exploring the Geometry and Portability of Sentiment Concept Vectors

Paper · 2026 · Laurits Lyngbaek, Pascale Feldkamp, Yuri Bizzoni et al.

Summary

Humanities research on sentiment, such as tracing emotional arcs in stories, often needs fine-grained, continuous sentiment scores. Lexicon tools handle longer context poorly, and Transformer classifiers mostly output discrete labels. The authors test Concept Vector Projection (CVP), which represents sentiment as a single direction in sentence-embedding space. To build the vector, they subtract the mean embedding of strongly negative sentences from the mean embedding of strongly positive ones. Each text is then scored by projecting its embedding onto this direction.

The study asks two questions: whether such vectors still work when the genre, era, language or emotional dimension changes, and whether the underlying assumption of linearity actually holds. Experiments use EmoBank, Facebook status updates and Fiction4 (English and Danish literature), all encoded with a multilingual sentence-transformer. Vectors trained on one corpus lose very little accuracy when applied to another. The method also works for arousal and dominance, though less reliably than for valence. A geometric analysis shows that neutral sentences sit off the positive–negative axis, so linearity is only approximate.

These findings suggest that historical or low-resource corpora may not require domain-specific training data.

Key points

  • Data: 19,257 human-annotated sentences in total from three corpora.
    • EmoBank: 10,062 sentences, Valence-Arousal-Dominance annotations.
    • Facebook: 2,895 posts, valence and arousal.
    • Fiction4: 6,300 lines/sentences, 1798–1965. Includes Plath, Hemingway, H.C. Andersen and Danish hymns.
  • Model: sentence-transformers/paraphrase-multilingual-mpnet-base-v2 (278M parameters).
  • Defining extremes: "positive" and "negative" are set at ±1 standard deviation from each corpus's mean valence.
  • Cross-corpus transfer: Spearman correlations with human valence range from 0.64 to 0.70. In-domain results are 0.66 (Fiction4), 0.70 (EmoBank) and 0.68 (Facebook).
  • Other affective dimensions (five-fold cross-validation):
    • EmoBank: valence .71, arousal .36, dominance .35.
    • Facebook: arousal .65.
  • Geometry: the centroids of the negative, neutral and positive classes form a triangle rather than a line. In Fiction4, the embeddings trace a curved, banana-shaped manifold. This indicates that neutral texts carry semantic content beyond simply lacking valence.

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

 Institutions

🎓Aarhus UniversityAarhus · Denmark6 works · 12 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksPsychology · 1,552 worksExperimental and Cognitive Psychology · 870 works

🧭 Topics 1Research subjects and languages the work is about.Sentiment & emotion · 972 works

Cite

Cite as Laurits Lyngbaek, Pascale Feldkamp, Yuri Bizzoni, Kristoffer Nielbo and Kenneth Enevoldsen. 2026. Is Sentiment Banana-Shaped? Exploring the Geometry and Portability of Sentiment Concept Vectors. The Proceedings for the 15th Workshop on Computational Approaches to Subjectivity, Sentiment Social Media Analysis (WASSA 2026).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.wassa-1.13

@inproceedings{lyngbaek2026is,
  title = {{Is Sentiment Banana-Shaped? Exploring the Geometry and Portability of Sentiment Concept Vectors}},
  author = {Laurits Lyngbaek, Pascale Feldkamp, Yuri Bizzoni, Kristoffer Nielbo and Kenneth Enevoldsen},
  booktitle = {The Proceedings for the 15th Workshop on Computational Approaches to Subjectivity, Sentiment Social Media Analysis (WASSA 2026)},
  year = {2026},
  doi = {10.18653/v1/2026.wassa-1.13},
  url = {https://doi.org/10.18653/v1/2026.wassa-1.13}
}

References

31 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. While the goal of this work, is not to explore how to optimize the performance of the CVP, but examine its implications, we only examine one model to ensure comparability with Lyngbaek et al. (2025). Further analysis should explore alternative models as indicate evidence suggests that newer models l… find
  2. Ali Allaith, Kirstine Degn, Alexander Conroy, Bolette Pedersen, Jens Bjerring-Hansen, and Daniel Hershcovich. 2023. Sentiment Classification of Historical Danish and Norwegian Literary Texts. In Proceedings of the 24th Nordic Conference on Computational Linguistics (NoDaLiDa), pages 324–334, Tórshav… find
  3. Francesco Barbieri, Luis Espinosa Anke, and Jose Camacho-Collados. 2022. XLM-T: Multilingual Language Models in Twitter for Sentiment Analysis and Beyond. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 258–266, Marseille, France. European Language Resources Asso…
  4. Vuk Batanović, Miloš Cvetanović, and Boško Nikolić. 2020. A versatile framework for resource-limited sentiment articulation, annotation, and analysis of short texts. PLoS ONE, 15(11). find
  5. Yuri Bizzoni and Pascale Feldkamp. 2023. Comparing transformer and dictionary-based sentiment models for literary texts: Hemingway as a case-study. In Proceedings of the 3rd International Workshop on Natural Language Processing for Digital Humanities, pages 219–226, Tokyo, Japan. Association for Com… find
  6. Yuri Bizzoni, Pascale Moreira, Mads Rosendahl Thomsen, and Kristoffer Nielbo. 2023. Sentimental matters - predicting literary quality by sentiment analysis and stylometric features. In Proceedings of the 13th Workshop on Computational Approaches to Subjectivity, Sentiment, & Social Media Analysis, p… find
  7. Sven Buechel and Udo Hahn. 2017. EmoBank: Studying the impact of annotation perspective and representation format on dimensional emotion analysis. In Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers, pages 578–585, Va… find
  8. Tobias Daudert. 2021. Exploiting textual and relationship information for fine-grained financial sentiment analysis. Knowledge-Based Systems, 230:107389. find
  9. Pascale Feldkamp, Ea Overgaard Lindhardt, Kristoffer L. Nielbo, and Yuri Bizzoni. 2024. Sentiment Below the Surface: Omissive and Evocative Strategies in Literature and Beyond. In Proceedings of the Computational Humanities Research Conference, volume 3834 of CEUR Workshop Proceedings, pages 681–706… find
  10. Matthew Jockers. 2014. A Novel Method for Detecting Plot. Matthew L. Jockers Blog. find
  11. Been Kim, Martin Wattenberg, Justin Gilmer, Carrie Cai, James Wexler, Fernanda Viegas, and 1 others. 2018. find
  12. Interpretability beyond feature attribution: Quantitative testing with concept activation vectors (tcav). In International conference on machine learning, pages 2668–2677. PMLR. find
  13. Michal Kosinski, David Stillwell, and Thore Graepel. 2013. Private traits and attributes are predictable from digital records of human behavior. Proceedings of the National Academy of Sciences, 110(15):5802–5805. find
  14. Belinda Z. Li, Maxwell Nye, and Jacob Andreas. 2021. Implicit representations of meaning in neural language models. In Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Pa… find
  15. Laurits Lyngbaek, Pascale Feldkamp, Yuri Bizzoni, Kristoffer L. Nielbo, and Kenneth Enevoldsen. 2025. Continuous Sentiment Scores for Literary and Multilingual Contexts. In Anthology of Computers and the Humanities, volume 3, pages 480–497, Luxembourg. ACH. find
  16. Albert Mehrabian and James A Russell. 1974. An approach to environmental psychology. the MIT Press. find
  17. Kiho Park, Yo Joong Choe, and Victor Veitch. 2024. The Linear Representation Hypothesis and the Geometry of Large Language Models. arXiv preprint. ArXiv:2311.03658 [cs].
  18. 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
  19. Andrew J Reagan, Lewis Mitchell, Dilan Kiley, Christopher M Danforth, and Peter Sheridan Dodds. 2016. The Emotional Arcs of Stories Are Dominated by Six Basic Shapes. EPJ Data Science, 5(1):1–12. find
  20. Simone Rebora, Marina Lehmann, Anne Heumann, Wei Ding, and Gerhard Lauer. 2023. Comparing Chat-GPT to human raters and sentiment analysis tools for german children’s literature. In Proceedings of the Computational Humanities Research Conference 2023, Paris, France, December 6-8, 2023, volume 3558 of… find
  21. Nils Reimers and Iryna Gurevych. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks. In Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics.
  22. Thomas Schmidt and Manuel Burghardt. 2018. An Evaluation of Lexicon-based Sentiment Analysis Techniques for the Plays of Gotthold Ephraim Lessing. In Proceedings of the Second Joint SIGHUM Workshop on Computational Linguistics for Cultural Heritage, Social Sciences, Humanities and Literature, pages… find
  23. Henrique Schechter Vera, Sahil Dua, Biao Zhang, Daniel Salz, Ryan Mullins, Sindhu Raghuram Panyam, Sara Smoot, Iftekhar Naim, Joe Zou, Feiyang Chen, Daniel Cer, Alice Lisak, Min Choi, Lucas Gonzalez, Omar Sanseviero, Glenn Cameron, Ian Ballantyne, Kat Black, Kaifeng Chen, and 70 others. 2025. Embedd… — [arXiv:2509.20354](https://arxiv.org/abs/2509.20354) find
  24. Krishnapriya Vishnubhotla, Adam Hammond, Graeme Hirst, and Saif Mohammad. 2024. The emotion dynamics of literary novels. In Findings of the Association for Computational Linguistics: ACL 2024, pages 2557–2574, Bangkok, Thailand. Association for Computational Linguistics. find
  25. Thuy Vu and D. Stott Parker. 2016. 𝑘-embeddings: Learning conceptual embeddings for words using context. In Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 1262–1267, San Diego, California. Associat… find
  26. Amy Beth Warriner, Victor Kuperman, and Marc Brysbaert. 2013. Norms of valence, arousal, and dominance for 13,915 English lemmas. Behavior Research Methods, 45(4):1191–1207. find
  27. Jan Wehner, Sahar Abdelnabi, Daniel Tan, David Krueger, and Mario Fritz. 2025. Taxonomy, Opportunities, and Challenges of Representation Engineering for Large Language Models. arXiv preprint. ArXiv:2502.19649 [cs]. — [arXiv:2502.19649](https://arxiv.org/abs/2502.19649) find
  28. Theresa Wilson, Janyce Wiebe, and Paul Hoffmann. 2005. Recognizing contextual polarity in phrase-level Sentiment Analysis. In Proceedings of Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing, pages 347–354, Vancouver, British Columbia, Canada. As…
  29. Haihua Xie and Miao He. 2025. Tracking Fine-Grained Public Opinions: Two Datasets from Online Discourse on Trending Topics. Mathematics, 13(21):3433. find
  30. Albin Zehe, Martin Becker, Lena Hettinger, Andreas Hotho, Isabella Reger, and Fotis Jannidis. 2016. Prediction of Happy Endings in German Novels Based on Sentiment Information. In Interactions between Data Mining and Natural Language Processing, pages 9–16, Riva del Garda. find
  31. Haiyan Zhao, Heng Zhao, Bo Shen, Ali Payani, Fan Yang, and Mengnan Du. 2024. Beyond single concept find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/lyngbaek-2026-is-sentiment-banana-shaped-exploring/