Integrated DAGCN-Bi-LSTM to Model Fairy Tale’s Character Relationships
JOIV : International Journal on Informatics Visualization · 30.03.2026 · 10 (2) · 712
DOI: 10.62527/joiv.10.2.4521 ↗Kurz gefasst
Fachartikel in der Zeitschrift JOIV : International Journal on Informatics Visualization, online erschienen am 30.03.2026. Verfasst von Erna Daniati und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).
Abstract (Originalsprache)
Story structure and literary meaning, at least in fairytales, are very much focused on the relationships between individual characters that reflect or communicate moral morals and self-hood. However, manual inspection and analysis of such dependencies remains subjective and arduous; it requires automation through computational methods. In this study, we introduce a novel method for automatic detection of character relations in Hans Christian Andersen’s fairy tales by the application of Dependency Attention-Aware Graph Convolutional Network (DAGCN)) and Bidirectional LSTM (Bi-LSTM). DAGCN encodes structural relationships using the syntactic dependency graph of a narrative, while Bi-LSTM optimizes both local and global context by encoding narrative sequence. Attention: We enhance the DAGCN by focusing on the more informative parts of dependency graph. Our model reached an F1 score of 92.1% on the Andersen corpus, surpassing the GCN-only and Bi-LSTM-only baselines. It could recover common types of relationships (e.g., love, conflict, friendship, and family) with high fidelity but not for less frequent and/or more subtle ones. An error analysis indicated that failure to capture context and rare character interactions were the leading causes of misclassification; these findings are clues for system improvement. Such a structure, more scalable, automatic and reliable than manual efforts, can be employed to analyze literature. Aside from Andersen, it proposes a way of treating character relations across literature. We envision future work on this that will extend it more fully to cover context, as well as be developed for multilingual/multimodal data-sets; in its wide applicability to computational narrative analysis we see it pushing forwards both the field of literary studies and AI more generally.
Zitieren
APA
Daniati, E., Wibawa, A. P., Irianto, W. S. G., & Nafalski, A. (2026). Integrated DAGCN-Bi-LSTM to Model Fairy Tale’s Character Relationships. JOIV : International Journal on Informatics Visualization, 10(2), 712. https://doi.org/10.62527/joiv.10.2.4521
Vancouver
Daniati E, Wibawa AP, Irianto WSG, Nafalski A. Integrated DAGCN-Bi-LSTM to Model Fairy Tale’s Character Relationships. JOIV : International Journal on Informatics Visualization. 2026;10(2):712. doi:10.62527/joiv.10.2.4521
BibTeX
@article{Daniati2026,
author = {Daniati, Erna and Wibawa, Aji Prasetya and Irianto, Wahyu Sakti Gunawan and Nafalski, Andrew},
title = {Integrated DAGCN-Bi-LSTM to Model Fairy Tale’s Character Relationships},
journal = {JOIV : International Journal on Informatics Visualization},
year = {2026},
volume = {10},
number = {2},
pages = {712},
doi = {10.62527/joiv.10.2.4521}
}
EUVIMED