An AI-Driven Computational Framework for Literary Analysis, Natural Language Processing, Semantic Interpretation, and Intelligent Decision Making in Digital Humanities.
Computer and Decision Making: An International Journal · 30.10.2026 · 3 (6S) · 3099-3108
DOI: 10.1922/comdem.v3i6s.19052 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computer and Decision Making: An International Journal, online erschienen am 30.10.2026. Verfasst von Dr. U. Aishwarya und 5 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
With the growth of available literary collections in digital format, new computational approaches are needed that are able to organize large collections while maintaining interpretive transparency. One of the challenges is that literary works can convey several similar themes, and standard multi-class classification is not sufficient and some pipelines for standard text analysis cannot be directly applied. This work proposes an Artificial Intelligence (AI) based computational framework which combines literary-text analysis, natural language processing, semantic representation, multi-label thematic prediction and interpretable decision support. The framework was tested on a project Gutenberg corpus of fiction books (6322) with 52 observed raw topic labels and 1716 authors of the books. Fourteen topics supported by 150 or more books were selected for confirmatory prediction resulting in a 6,249 book modelling set. Books were segmented by author, so that no authors were represented in all three data partitions (training, validation and test). For comparison, a TF-IDF one-vs-rest logistic-regression model was tested alongside a model based on semantic embedding of sentences (Sentence-BERT), using the decision thresholds selected in the validation set. The TF-IDF model performed better on the held-out test set with macro-F1 scores of 0.519, micro-F1 scores of 0.711 and weighted-F1 scores of 0.716 compared to the semantic model results of 0.428, 0.586 and 0.631. Semantic clustering resulted in poor separation (best Cosine silhouette coefficient: 0.078; two clusters), and a high amount of thematic overlap. The framework integrates the topic probabilities, the lexical indicators that humans can interpret and the evidence from the semantic neighbors to complement, not supplant, human literary judgement. The results demonstrate that transparent lexical representations are still very efficient for large scale literary classification and semantic representations add complementary contextual evidence for decision support in the digital humanities.
Zitieren
APA
Aishwarya, D. U., Karun, D. K. G., V, D. S., Tomar, D. Y., Pavithra, D. M., & Malaviya, D. R. (2026). An AI-Driven Computational Framework for Literary Analysis, Natural Language Processing, Semantic Interpretation, and Intelligent Decision Making in Digital Humanities. Computer and Decision Making: An International Journal, 3(6S), 3099-3108. https://doi.org/10.1922/comdem.v3i6s.19052
Vancouver
Aishwarya DU, Karun DKG, V DS, Tomar DY, Pavithra DM, Malaviya DR. An AI-Driven Computational Framework for Literary Analysis, Natural Language Processing, Semantic Interpretation, and Intelligent Decision Making in Digital Humanities. Computer and Decision Making: An International Journal. 2026;3(6S):3099-3108. doi:10.1922/comdem.v3i6s.19052
BibTeX
@article{Aishwarya2026,
author = {Aishwarya, Dr. U. and Karun, Dr. K. Gayathri and V, Dr. Sutharshan and Tomar, Dr. Yashika and Pavithra, Dr. M and Malaviya, Dr. Richa},
title = {An AI-Driven Computational Framework for Literary Analysis, Natural Language Processing, Semantic Interpretation, and Intelligent Decision Making in Digital Humanities},
journal = {Computer and Decision Making: An International Journal},
year = {2026},
volume = {3},
number = {6S},
pages = {3099--3108},
doi = {10.1922/comdem.v3i6s.19052}
}
EUVIMED