Elamur Research⌕
☰
home / papers / 2025 / Narrative Conflicts: A Tri-Modal Computational Analysis of Antagonism in Shakespeare's Julius Caesar

Narrative Conflicts: A Tri-Modal Computational Analysis of Antagonism in Shakespeare's Julius Caesar

Paper · 2025 · Mehmet Can Yavuz, Lucia Cascone, Aylin Özkan et al.

Summary

The paper asks whether computational tools can expose the different kinds of conflict in a play and what they reveal about its structure. The authors treat antagonism as three layers: clashing ideas (semantic), clashing feelings (emotional), and clashing social positions (relational). They test this on Shakespeare's Julius Caesar, using each character's speeches as the unit of analysis. For the semantic layer, they average speech embeddings per character, compare characters by cosine distance, group them with Ward hierarchical clustering, and plot them with t-SNE. For the emotional layer, they cluster characters on both the mean and the variance of their emotion classifier outputs. For the relational layer, they build a scene co-occurrence graph and measure degree and betweenness. The semantic clusters track dramatic roles. Emotional variance separates volatile characters from steady ones. Unexpectedly, minor figures such as the Servant and Lucius are the main bridges in the network. The authors connect these results to Hegelian and Nietzschean readings, arguing that AI tools can support interpretive literary criticism.

Key points

  • Each character is represented by the mean of the encoder vectors of their speeches. Pairwise distance is 1 minus cosine similarity, followed by Ward-linkage clustering and t-SNE projection.
  • Semantic clusters line up with narrative function. Conspirators and statesmen form a core group, while servants and battlefield figures are outliers. Popilius Lena is isolated as a "semantic singularity."
  • In the emotional analysis, mean profiles yield fearful, angry and neutral archetypes. Variance profiles single out Decius Brutus and Marullus as high-variance oscillators, while Calpurnia is emotionally static.
  • The co-occurrence network has four groups:
    • a citizenry clique,
    • a conspiratorial circle (Brutus, Cassius, Casca, Trebonius and others),
    • a post-assassination military-political group,
    • peripheral actors such as Lepidus and Cicero.
  • In the same network, the Servant and Lucius act as high-betweenness hubs, with Messala as a secondary hub.
  • The authors list two limitations. The emotion resources are modern and may misread Early Modern English. Characters with few lines get unstable embeddings.

Interesting ideas

  • I1. Cluster on the variance of a signal, not only its mean, to separate characters who swing between emotions from those who stay fixed. Averages alone hide this distinction.
  • I2. Measure betweenness centrality on a character co-occurrence graph to find low-prominence mediators that frequency counts would rank as unimportant.
  • I3. Split one abstract notion, conflict, into separate measurable views and then check where they agree. Here, characters central in the network tend to sit between semantic clusters and show moderate emotional variance.
  • I4. Flag a data-sparsity warning: entities with very few utterances yield unreliable aggregate representations.
  • I5. Propose a sliding window across acts as future work, so that relations and stances are tracked over time rather than as one global snapshot.

For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

  • I1 (frames): The prag crate could track per-entity variance of speech-act or sentiment labels across a story, flagging characters whose stance shifts as candidates for irony or deception.
  • I2 (world): Building a scene-level interaction graph over world entities would let Mova explain why a minor messenger or servant is causally important to the plot.
  • I3 (answer): For QA about "who opposes whom" in fables, Mova could combine role-based, affect-based and interaction-based evidence and cite each channel in its explanation.
  • I4 (learn): Evaluation should report results bucketed by how much text each entity has, and withhold confident judgments about rarely mentioned characters.
  • I5 (world): Mova's event ordering could record states per episode, so the world model shows how alliances and hostilities change instead of collapsing them into one static relation.

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, topics and concepts

📚 Shelves 2The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 works

Cite

Cite as Mehmet Can Yavuz, Lucia Cascone, Aylin Özkan and İrem Ertaş. 2025. Narrative Conflicts: A Tri-Modal Computational Analysis of Antagonism in Shakespeare's Julius Caesar. Proceedings of the Eleventh Italian Conference on Computational Linguistics (CLiC-it 2025).

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{yavuz2025narrative,
  title = {{Narrative Conflicts: A Tri-Modal Computational Analysis of Antagonism in Shakespeare's Julius Caesar}},
  author = {Mehmet Can Yavuz, Lucia Cascone, Aylin Özkan and İrem Ertaş},
  booktitle = {Proceedings of the Eleventh Italian Conference on Computational Linguistics (CLiC-it 2025)},
  year = {2025},
  url = {https://aclanthology.org/2025.clicit-1.80/}
}

References

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

  1. I. Mani, Computational modeling of narrative, books.google.com, 2022. URL: https://books.google.com/books?hl=en&lr=&id=w4hyEAAAQBAJ& oi=fnd&pg=PP1&dq=advanced+computational+ tools+probe+the+complexities+of+narrative+ conflict+in+dramatic+texts&ots=JbD3gjWNUj& sig=LeDYu9IlBQwJPvyZbVlgvncEbyw. ↗
  2. P. Ranade, S. Dey, A. Joshi, T. Finin, Computational understanding of narratives: A survey, IEEE Access 10 (2022) 119872–119888. URL: https://ieeexplore.ieee.org/abstract/document/9882117/. doi: https://ieeexplore.ieee.org/ abstract/document/9882117/ . ↗
  3. M. Escobar Varela, Theater as data: Computational journeys into theater research, University of Michigan Press, 2021. find
  4. M. Riedl, R. Young, Narrative planning: Balancing plot and character, Journal of Artificial Intelligence Research 39 (2010) 1–40. URL: https://www.jair.org/ index.php/jair/article/view/10669. ↗
  5. S. Kusal, S. Patil, J. Choudrie, K. Kotecha, D. Vora, I. Pappas, A systematic review of applications of natural language processing and future challenges with special emphasis in text-based emotion detection, Artificial Intelligence Review 56 (2023) 15129–15215. find
  6. S. Anwar, M. O. Beg, K. Saleem, Z. Ahmed, Social relationship analysis using state-of-the-art embeddings, Information Processing (2023). URL: https://dl.acm.org/doi/abs/10.1145/3539608. doi: 10. 1145/3539608 . ↗
  7. M. Grandjean, Network visualization: mapping shakespeare’s tragedies (2015). find
  8. M. E. Hartsock, The complexity of julius caesar, PMLA 81 (1966) 777–799. URL: https://www.cambridge.org/core/journals/ pmla/article/complexity-of-julius-caesar/ 3FA05E829345562E33A0F8B8371972D7. ↗
  9. E. Koomson, Rhetoric and political power in the last century of the Roman Republic, Ph.D. thesis, University of Cape Coast, 2020. find
  10. B. Leiter, Nietzsche’s moral and political philosophy, 2004. URL: https://plato.stanford.edu/ENTRIES/ nietzsche-moral-political/. ↗
  11. S. O’Dair, Social role and the making of identity in julius caesar, Studies in English Literature, 1500-1900 33 (1993) 289–307. find
  12. M. C. Yavuz, Analyses of literary texts by using statistical inference methods., in: CLiC-it, 2019. find
  13. M. C. Yavuz, Analyses of character emotions in dramatic works by using emolex unigrams (2020). find
  14. M. C. Yavuz, Analyses of dramatic network simulations by using markov chains, in: 2021 8th International Conference on Behavioral and Social Computing (BESC), 2021, pp. 1–6. find
  15. O. A. Rosso, H. Craig, P. Moscato, Shakespeare and other english renaissance authors as characterized by information theory complexity quantifiers, Physica A: Statistical Mechanics and its Applications 388 (2009) 916–926. find
  16. E. Kim, R. Klinger, A survey on sentiment and emotion analysis for computational literary studies, Digital Humanities Quarterly 13 (2019). find
  17. H. Makhdom, R. Li, H. F. Wu, K. Lui, S. Zarrieß, A review on sentiment and emotion analysis for computational literary studies, arXiv preprint arXiv:2402.18673 (2024). ↗
  18. T. Schmidt, K. Dennerlein, C. Wolff, Towards a corpus of historical german plays with emotion annotations, in: 3rd Conference on Language, Data and Knowledge (LDK 2021), Schloss-Dagstuhl-Leibniz Zentrum für Informatik, 2021. find
  19. T. Schmidt, K. Dennerlein, C. Wolff, Results of emotion annotation in german drama from 1650-1815., in: DH, 2023. find
  20. L. Christ, S. Amiriparian, M. Milling, I. Aslan, B. W. Schuller, Automatic emotion modelling in written stories, arXiv preprint arXiv:2212.11382 (2022). ↗
  21. J. P. Dexter, T. Katz, N. Tripuraneni, T. Dasgupta, A. Kannan, J. A. Brofos, J. A. Bonilla Lopez, L. A. Schroeder, A. Casarez, M. Rabinovich, A. Haimson Lushkov, P. Shah, Quantitative criticism of literary relationships, Proceedings of the National Academy of Sciences 114 (2017) E3195–E3204. find
  22. V. Perri, L. Qarkaxhija, A. Zehe, A. Hotho, I. Scholtes, One graph to rule them all: Using nlp and graph neural networks to analyse tolkien’s legendarium, arXiv preprint arXiv:2210.07871 (2022). ↗
  23. H. O. Hatzel, B. Krautter, N. Reiter, Machine learning in computational literary studies, Journal of Computational Literary Studies 2 (2023). find
  24. Y. Yin, Y. Wang, T. Ding, C. Shi, Beyond canonical texts: A computational analysis of fanfiction, arXiv preprint arXiv:2008.11518 (2020). ↗
  25. M. P. Eve, J. B. Herrmann, A. Piper, Computational text analysis within the humanities, Digital Humanities Quarterly 13 (2019). find
  26. M. Kestemont, J. Stover, M. Koppel, F. Karsdorp, W. Daelemans, Authenticating the writings of julius caesar, Expert Systems with Applications 63 (2016) 86–96. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/yavuz-2025-narrative-conflicts-tri-modal-computational/