---
title: "Tracing Thematic Change in Early English-Language Science"
description: "To counter this, the author builds a corpus of 238 public-domain Project Gutenberg books with LLM-inferred publication years… 20 references."
authors: ["Jonathan Gordon"]
year: 2026
venue: "Proceedings of the 6th International Conference on Natural Language Processing for the Digital Humanities"
source: "https://aclanthology.org/2026.nlp4dh-1.21/"
doi: "10.18653/v1/2026.nlp4dh-1.21"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/gordon-2026-tracing-thematic-change-early-english/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/gordon-2026-tracing-thematic-change-early-english/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/46/) / Tracing Thematic Change in Early English-Language Science Fiction, 1818-1930

# Tracing Thematic Change in Early English-Language Science Fiction, 1818-1930

**Authors:** Jonathan Gordon · **Year:** 2026 · **Venue:** [Proceedings of the 6th International Conference on Natural Language Processing for the Digital Humanities](https://md.research.elamur.ai/free/en/venues/nlp4dh/) ([International Conference on Natural Language Processing for the Digital Humanities 2026](https://md.research.elamur.ai/free/en/venues/nlp4dh/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.nlp4dh-1.21/) · [PDF](https://aclanthology.org/2026.nlp4dh-1.21.pdf) · [DOI 10.18653/v1/2026.nlp4dh-1.21](https://doi.org/10.18653/v1/2026.nlp4dh-1.21) · [OpenAlex](https://openalex.org/W7166823807)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
The paper investigates how the range of themes in early science fiction written in English changed from Frankenstein (1818) to the early pulp era (1930). Topic models on literary collections tend to pick up authorship rather than themes, especially when a corpus is small and dominated by a few writers. To counter this, the author builds a corpus of 238 public-domain Project Gutenberg books with LLM-inferred publication years, cuts them into roughly 1,000-word chunks, and trains a separate MALLET LDA model for each of four time bins. Each binned model is run with and without the Authorless transformation, which suppresses author-specific vocabulary. Topics in adjacent bins are then linked by similarity and tested against a permutation null.

The links between periods beat chance. Authorless spreads topics across more authors and slightly raises the share of thematic topics, with no real loss of coherence. Four persistent chains emerge, around mobility, affect, planetary scale and scientific knowledge. They suggest a movement from romantic speculation toward technoscience. The workflow is useful for diachronic study of small, author-skewed collections.

## Key points
- **Corpus.** 238 books and 14,973 chunks drawn from the Hugging Face dataset stevez80/Sci-Fi-Books-gutenberg. Years come from Momen et al. (2025). A spot check found 17 of 20 dates within ±5 years.
- **Bins.** Pre-1880, 1880–1899, 1900–1914 and 1915–1930. Jules Verne supplies 42.5% of chunks in the earliest bin, and 67% of books date from 1900–1930. Only 9 of 25 well-known early SF works were found in the source data.
- **Why bin at all.** Whole-corpus LDA was unstable, with mean Jaccard of 0.26–0.31 across seeds. A larger full-corpus model (k = 60) reaches similar topic quality but cannot support alignment across periods.
- **Setup.**
  - MALLET, 1,000 iterations, k = 15 per bin.
  - Document-frequency filter keeping words in ≥10 chunks and ≤50% of chunks, plus a 178-word stoplist, with no lemmatization.
  - Names removed with a capitalization heuristic.
  - Authorless implemented via DownsampleLabelWords with threshold 0.05.
- **Evaluation.** 195 topic instances across four configurations, judged manually on three axes: coherence, thematic content and author concentration. Author concentration is the share of a topic's top-5 chunks that come from its dominant author. Alignment uses cosine similarity on L2-normalized topic-word distributions.

## Interesting ideas
- Quality is reported on separate axes (does the topic hang together, is it about subject matter, is it one author's idiom) instead of a single score, which shows what each intervention actually changes.
- Cross-period topic matches are checked against a permutation null, so continuity has to beat what random pairings would produce.
- A simple diagnostic, the fraction of a pattern's strongest examples that come from one source, flags patterns that reflect an idiosyncratic voice rather than a general regularity.
- Downsampling tokens that are strongly tied to a source label, before computing statistics, reduces how much a dominant contributor shapes the result.
- The author notes that bag-of-words topics miss plot-level and relational themes such as time travel or first contact.

## For Mova

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

- **I1** (learn): evaluation of parser or frame output could report agreement, semantic validity and source concentration separately, so a fix that helps one axis is not mistaken for overall progress.
- **I2** (learn): agreement among witness parsers on silver data could be compared with a shuffled-pairing baseline as a built-in negative control.
- **I3** (world): before adding induced verb-class or ontology facts to compiled knowledge, check whether their supporting examples come mostly from a single book or author.
- **I4** (tokens): lexicon and tagger frequency counts built from fiction could downweight book-specific words, so that one prolific author (a Verne equivalent) does not skew lemma or tag priors.
- **I5** (world): Mova's event order and who-did-what-to-whom frames could track such themes directly, which shows where a structured world model goes beyond topic modeling.

*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

- [Vassar College](https://md.research.elamur.ai/free/en/universities/vassar-college/) — University, Town of Poughkeepsie, United States · 6 works · 1,118 citations

📚 Shelves (6): [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works · [🤝 Social Sciences](https://md.research.elamur.ai/free/en/shelves/social-sciences/) · 1,772 works · [🤔 Philosophy](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/philosophy/) · 284 works · [🤝 General Social Sciences](https://md.research.elamur.ai/free/en/shelves/social-sciences/general-social-sciences/) · 227 works · [📚 Literature and Literary Theory](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/literature-and-literary-theory/) · 519 works

🧭 Topics (1): [English](https://md.research.elamur.ai/free/en/topics/english/) · 2,263 works

## Cite

Cite as: Jonathan Gordon. 2026. Tracing Thematic Change in Early English-Language Science Fiction, 1818-1930. Proceedings of the 6th International Conference on Natural Language Processing for the Digital Humanities.

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

```bibtex
@inproceedings{gordon2026tracing,
  title = {{Tracing Thematic Change in Early English-Language Science Fiction, 1818-1930}},
  author = {Jonathan Gordon},
  booktitle = {Proceedings of the 6th International Conference on Natural Language Processing for the Digital Humanities},
  year = {2026},
  doi = {10.18653/v1/2026.nlp4dh-1.21},
  url = {https://doi.org/10.18653/v1/2026.nlp4dh-1.21}
}
```

- [← previous Tracing the Light of Thought: A Probabilistic Self- and Cross-Consistency Verification Mechanism Improving Mathematical Reasoning in LLMs](https://md.research.elamur.ai/free/en/papers/liu-2026-tracing-light-thought-probabilistic-self/)
- [next → Tracking Autism Stigma in Italian Newspapers: A Longitudinal Analysis of Media Discourse (2016–2025)](https://md.research.elamur.ai/free/en/papers/martinelli-2026-tracking-autism-stigma-italian-newspapers/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/46/)

## References

20 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. This paper shows that temporally binned topic modeling can support diachronic analysis in a small, author-skewed literary corpus. A comparison of four model configurations finds that full-corpus LDA at adequate capacity produces comparable per-topic quality, but only binned models enable cross-period alignment. Within that setting, Authorless preprocessing reduces author concentration and modestly increases thematicity without materially reducing coherence. Applied to early English-language science fiction (1818–1930), the workflow yields four high-continuity chains – cen-Maria Antoniak. 2022. Topic modeling for the people. Blog post. [find](https://scholar.google.com/scholar?q=This+paper+shows+that+temporally+binned+topic+modeling+can+support+diachronic+analysis+in+a+small%2C+author%2Dskewed+literary+corpus%2E+A+comparison+of+four+model+configurations+finds+that+full%2Dcorpus+LDA+a)
2. David M. Blei, Andrew Y. Ng, and Michael I. Jordan. 2003. [Latent Dirichlet allocation](https://md.research.elamur.ai/free/en/papers/blei-2003-latent-dirichlet-allocation/). Journal of Machine Learning Research, 3:993–1022. [↗](https://doi.org/10.5555/944919.944937)
3. Federica Bologna. 2020. A computational approach to urban space in science fiction. Journal of Cultural Analytics, 5(2). [↗](https://doi.org/10.22148/001c.18120)
4. Jonathan Chang, Sean Gerrish, Chong Wang, Jordan Boyd-Graber, and David Blei. 2009. [Reading tea leaves: How humans interpret topic models](https://md.research.elamur.ai/free/en/papers/chang-2009-reading-tea-leaves-humans-interpret-topic/). In Advances in Neural Information Processing Systems, volume 22. Curran Associates, Inc.
5. John Clute, David Langford, and Peter Nicholls. 2021. The encyclopedia of science fiction. Ansible Editions. Accessed 2025. the Association for Computational Linguistics: Volume 2, Short Papers, pages 432–436, Valencia, Spain. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=John+Clute%2C+David+Langford%2C+and+Peter+Nicholls%2E+2021%2E+The+encyclopedia+of+science+fiction%2E+Ansible+Editions%2E+Accessed+2025%2E+the+Association+for+Computational+Linguistics%3A+Volume+2%2C+Short+Papers%2C+pages)
6. Caitlin Doogan and Wray Buntine. 2021. Topic model or topic twaddle? Re-evaluating semantic interpretability measures. In Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 3824–3848, Online. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Caitlin+Doogan+and+Wray+Buntine%2E+2021%2E+Topic+model+or+topic+twaddle%3F+Re%2Devaluating+semantic+interpretability+measures%2E+In+Proceedings+of+the+2021+Conference+of+the+North+American+Chapter+of+the+Associ)
7. Alexandra Schofield and David Mimno. 2016. Comparing apples to apple: The effects of stemmers on topic models. Transactions of the Association for Computational Linguistics, 4:287–300. [↗](https://doi.org/10.1162/tacl_a_00099)
8. Derek Greene, Derek O’Callaghan, and Pádraig Cunningham. 2014. How many topics? stability analysis for topic models. In Machine Learning and Knowledge Discovery in Databases, pages 498–513, Berlin, Heidelberg. Springer. [↗](https://doi.org/10.48550/arxiv.1404.4606)
9. Alexander Hoyle, Lorena Calvo-Bartolomé, Jordan Boyd-Graber, and Philip Resnik. 2025. ProxAnn: Use-oriented evaluations of topic models and document clustering. In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 15872– 15897, Vienna, Austria. Association for Computational Linguistics. [↗](https://doi.org/10.48550/arxiv.2507.00828)
10. Oleg Sobchuk and Artjoms Šel , a. 2024. Computational thematics: Comparing algorithms for clustering the genres of literary fiction. Humanities and Social Sciences Communications, 11:1–12. [↗](https://doi.org/10.1057/s41599-024-02933-6)
11. Laure Thompson and David Mimno. 2018. Authorless topic models: Biasing models away from known structure. In Proceedings of the 27th International Conference on Computational Linguistics, pages 3903–3914, Santa Fe, New Mexico, USA. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Laure+Thompson+and+David+Mimno%2E+2018%2E+Authorless+topic+models%3A+Biasing+models+away+from+known+structure%2E+In+Proceedings+of+the+27th+International+Conference+on+Computational+Linguistics%2C+pages+3903%E2%80%9339)
12. Joris J. van Zundert, Marijn Koolen, Julia Neugarten, Peter Boot, Willem van Hage, and Ole Mussmann. 2022. What do we talk about when we talk about topic? In Computational Humanities Research Conference, pages 1–18. [find](https://scholar.google.com/scholar?q=Joris+J%2E+van+Zundert%2C+Marijn+Koolen%2C+Julia+Neugarten%2C+Peter+Boot%2C+Willem+van+Hage%2C+and+Ole+Mussmann%2E+2022%2E+What+do+we+talk+about+when+we+talk+about+topic%3F+In+Computational+Humanities+Research+Conferen)
13. Alexander Hoyle, Pranav Goel, Andrew Hian-Cheong, Denis Peskov, Jordan Boyd-Graber, and Philip Resnik. 2021. Is automated topic model evaluation broken? the incoherence of coherence. In Advances in Neural Information Processing Systems, volume 34, pages 2018–2033. Curran Associates, Inc. [↗](https://doi.org/10.48550/arxiv.2107.02173)
14. Matthew L. Jockers. 2013. Macroanalysis: Digital Methods and Literary History. University of Illinois Press. [find](https://scholar.google.com/scholar?q=Matthew+L%2E+Jockers%2E+2013%2E+Macroanalysis%3A+Digital+Methods+and+Literary+History%2E+University+of+Illinois+Press%2E)
15. Matthew L. Jockers and David Mimno. 2013. Significant themes in 19th-century literature. Poetics, 41(6):750–769. [↗](https://doi.org/10.1016/j.poetic.2013.08.005)
16. Andrew Kachites McCallum. 2002. MALLET: A machine learning for language toolkit. [find](https://scholar.google.com/scholar?q=Andrew+Kachites+McCallum%2E+2002%2E+MALLET%3A+A+machine+learning+for+language+toolkit%2E)
17. 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 Linguistics. [find](https://scholar.google.com/scholar?q=Omar+Momen%2C+Manuel+Schaaf%2C+and+Alexander+Mehler%2E+2025%2E+Filling+the+temporal+void%3A+Recovering+missing+publication+years+in+the+Project+Gutenberg+corpus+using+LLMs%2E+In+Findings+of+the+Association+for+Co)
18. Lisa M. Rhody. 2013. Topic modeling and figurative language. Journal of Digital Humanities, 2(1). [find](https://scholar.google.com/scholar?q=Lisa+M%2E+Rhody%2E+2013%2E+Topic+modeling+and+figurative+language%2E+Journal+of+Digital+Humanities%2C+2%281%29%2E)
19. Benjamin M. Schmidt. 2013. Words alone: Dismantling topic models in the humanities. Journal of Digital Humanities, 2(1). [find](https://scholar.google.com/scholar?q=Benjamin+M%2E+Schmidt%2E+2013%2E+Words+alone%3A+Dismantling+topic+models+in+the+humanities%2E+Journal+of+Digital+Humanities%2C+2%281%29%2E)
20. Alexandra Schofield, Måns Magnusson, and David Mimno. 2017. Pulling out the stops: Rethinking stopword removal for topic models. In Proceedings of the 15th Conference of the European Chapter of [find](https://scholar.google.com/scholar?q=Alexandra+Schofield%2C+M%C3%A5ns+Magnusson%2C+and+David+Mimno%2E+2017%2E+Pulling+out+the+stops%3A+Rethinking+stopword+removal+for+topic+models%2E+In+Proceedings+of+the+15th+Conference+of+the+European+Chapter+of)

*Updated 1 November 2026*