Elamur Research⌕
☰
home / papers / 2026 / Predicting Topic (Co-)Occurrence Using Topic Networks Built from the Project Gutenberg Corpus

Predicting Topic (Co-)Occurrence Using Topic Networks Built from the Project Gutenberg Corpus

Summary

Temporal topic analysis has mostly targeted scientific and legal writing. This paper applies it to literature and asks whether the thematic structure of past books can forecast future themes and how they combine. The authors use the temporal Project Gutenberg corpus of Momen et al. (2025), whose publication years were estimated with LLMs and RAG. From it they take PG [1700,1920], a subcorpus of books from 1700 to 1920, chosen to avoid sparsely populated periods.

Each book receives topics from supervised classifiers based on word embeddings and Transformers. Three topic systems are used, including the PG bookshelf categories and Dewey Decimal Classification (DDC). For each time span, the authors build a network in which topics are nodes and edges are weighted by how many books share both topics. They first measure temporal autocorrelation at the node and graph level. They then train models on structural features to predict which topics (nodes) and which topic pairings (edges) will appear in the next span.

Prediction quality is consistently high, which suggests that literary themes evolve in a fairly predictable way.

Key points

  • Data: In total, 72,978 books make up the temporal PG corpus. Of these, 53,774 dated between 1600 and 2000 are treated as reliably dated, and the study uses the 1700–1920 subset.
  • Topic systems: These include the Bookshelf Category System (BCS), with 394 categories, and DDC labels assigned by a text2ddc-based model (Baumartz, 2020).
  • Results: Across all three topic systems, ROC-AUC and Average Precision for node and link prediction stay above 0.85.
  • Effect of span length:
    • Longer spans give more stable predictions and better recall of which topics will be present.
    • Shorter spans capture changing relationships between topics more effectively.
  • Limitations: Node prediction is limited to topics that already exist, and the models predict only whether topics and links are present, not their strength. However, strong autocorrelation of node degree and average degree suggests that strength could also be forecast.

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 10.63317/58x3h7gjbpb4

@inproceedings{verma2026predicting,
  title = {{Predicting Topic (Co-)Occurrence Using Topic Networks Built from the Project Gutenberg Corpus}},
  author = {Verma, Bhuvanesh and Mehler, Alexander},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/58x3h7gjbpb4},
  url = {https://doi.org/10.63317/58x3h7gjbpb4}
}

References

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

  1. Ayman Al-Kababji, Faycal Bensaali, and Sarada Prasad Dakua. 2022. Scheduling techniques for liver segmentation: Reducelronplateau vs onecyclelr. In International conference on intelligent systems and pattern recognition, pages 204–212. Springer. find
  2. Daniel Baumartz. 2020. Automatic Topic Modeling in the Context of Digital Libraries: Mehrsprachige Korpus-basierte Erweiterung von text2ddc - eine experimentelle Studie. find
  3. Saman Behrouzi, Zahra Shafaeipour Sarmoor, Khosrow Hajsadeghi, and Kaveh Kavousi. 2020. Predicting scientific research trends based on link prediction in keyword networks. Journal of Informetrics, 14(4):101079. find
  4. Charlene Chou and Tony Chu. 2022. An analysis of bert (nlp) for assisted subject indexing for project gutenberg. Cataloging & Classification Quarterly, 60(8):807–835. find
  5. Nazim Choudhury, Fahim Faisal, and Matloob Khushi. 2020. Mining temporal evolution of knowledge graphs and genealogical features for literature-based discovery prediction. Journal of Informetrics, 14(3):101057. find
  6. Nazim Choudhury and Shahadat Uddin. 2016. Time-aware link prediction to explore network effects on temporal knowledge evolution. Scientometrics, 108(2):745–776. find
  7. Melvil Dewey. 1876. A classification and subject index, for cataloguing and arranging the books and pamphlets of a library. Brick row book shop, Incorporated. find
  8. Mattia Egloff, Davide Picca, and Alessandro Adamou. 2019. Extraction of character profiles from the gutenberg archive. In Research Conference on Metadata and Semantics Research, pages 367–372. Springer. find
  9. Martin Gerlach and Francesc Font-Clos. 2020. A standardized project gutenberg corpus for statistical analysis of natural language and quantitative linguistics. Entropy, 22(1):126. find
  10. Vasilii A Gromov and Quynh Nhu Dang. 2023. Semantic and sentiment trajectories of literary masterpieces. Chaos, Solitons & Fractals, 175:113934. find
  11. Hawoong Jeong, Zoltan Néda, and Albert-László Barabási. 2003. Measuring preferential attachment in evolving networks. Europhysics letters, 61(4):567. find
  12. Ming Jiang, Yuerong Hu, Glen Worthey, Ryan C Dubnicek, Boris Capitanu, Deren Kudeki, and J Stephen Downie. 2021. The gutenberg-hathitrust parallel corpus: A real-world dataset for noise investigation in uncorrected ocr texts. find
  13. Yookyung Jo, John E Hopcroft, and Carl Lagoze. 2011. The web of topics: discovering the topology of topic evolution in a corpus. In Proceedings of the 20th international conference on World wide web, pages 257–266. find
  14. Sukhwan Jung and Aviv Segev. 2022. Analyzing the generalizability of the network-based topic emergence identification method. Semantic Web, 13(3):423–439. find
  15. Taja Kuzman and Nikola Ljubešić. 2025. Llm teacher-student framework for text classification with no manually annotated data: A case study in iptc news topic classification. IEEE Access, pages 1–1. find
  16. Weibin Lin, Xianli Wu, Zhengwei Wang, Xiaoji Wan, and Hailin Li. 2022. Topic network analysis based on co-occurrence time series clustering. Mathematics, 10(16):2846. find
  17. Jean-Baptiste Michel, Yuan Kui Shen, Aviva Presser Aiden, Adrian Veres, Matthew K Gray, Google Books Team, Joseph P Pickett, Dale Hoiberg, Dan Clancy, Peter Norvig, et al. 2011. Quantitative analysis of culture using millions of digitized books. science, 331(6014):176–182. find
  18. Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781. find
  19. Omar Momen, Manuel Schaaf, and Alexander Mehler. 2025. Filling the temporal void: Recovering missing publication years in the Project Gutenberg corpus using LLMs. In Findings of the Association for Computational Linguistics: ACL 2025, pages 17318–17334, Vienna, Austria. Association for Computational… find
  20. Elaheh Momeni, Shanika Karunasekera, Palash Goyal, and Kristina Lerman. 2018. Modeling evolution of topics in large-scale temporal text corpora. In Proceedings of the International AAAI Conference on Web and Social Media, volume 12. find
  21. Mark E. J. Newman. 2010. Networks: An Introduction. Oxford University Press, Oxford. find
  22. Eitan Adam Pechenick, Christopher M Danforth, and Peter Sheridan Dodds. 2015. Characterizing the google books corpus: Strong limits to inferences of socio-cultural and linguistic evolution. PloS one, 10(10):e0137041. find
  23. Jean M Twenge, W Keith Campbell, and Brittany Gentile. 2012. Increases in individualistic words and phrases in american books, 1960–2008. PloS one, 7(7):e40181. find
  24. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. Attention is all you need. Advances in neural information processing systems, 30. find
  25. Yuan Yao, Lorenzo Rosasco, and Andrea Caponnetto. 2007. On early stopping in gradient descent learning. Constructive approximation, 26(2):289–315. find
  26. Jeremy York. 2009. This library never forgets: Preservation, cooperation, and the making of hathitrust digital library. In Archiving Conference, volume 6, pages 5–10. Society of Imaging Science and Technology. 10. find
  27. Language Resource References Momen, Omar and Schaaf, Manuel and Mehler, Alexander. 2025. Filling the Temporal Void: Recovering Missing Publication Years in the Project Gutenberg Corpus Using LLMs. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/verma-2026-predicting-topic-co-occurrence-topic/