Stevenson Text Analysis
Digital humanities · corpus methods · close-reading support
A reproducible digital-humanities demonstration using public-domain Robert Louis Stevenson texts. The project pairs literary research with lightweight computational methods while treating quantitative signals as prompts for close reading rather than substitutes for interpretation.
The initial corpus uses Project Gutenberg editions of The Strange Case of Dr. Jekyll and Mr. Hyde and Treasure Island.
What the project demonstrates
- Reproducible acquisition of open textual data
- Source documentation and corpus provenance
- Text cleaning and tokenization
- Descriptive corpus comparison
- Lexical-diversity measures
- Exploratory keyword analysis
- Responsible integration of quantitative and humanistic methods
Reproducibility
The repository retrieves the two documented Project Gutenberg sources at run time, validates the corpus manifest, cleans and tokenizes the text, and generates descriptive comparison output.
A GitHub Actions workflow now reruns the analysis automatically on pushes and pull requests. The first live-source validation completed successfully after the workflow was added.
The computational measures are deliberately framed as exploratory evidence for further reading, not as stand-alone interpretations of Stevenson’s texts.
Corpus comparison
| Measure | Jekyll & Hyde | Treasure Island |
|---|---|---|
| Word count | 25,976 | 70,498 |
| Unique tokens | 3,923 | 5,903 |
| Type-token ratio | 0.151 | 0.084 |
| Average token length | 4.14 | 3.91 |
Corpus size
The Strange Case of Dr. Jekyll and Mr. Hyde — 25,976 words
Treasure Island — 70,498 words
Selected term counts
| Term | Jekyll & Hyde | Treasure Island |
|---|---|---|
| body | 21 | 20 |
| face | 40 | 52 |
| hand | 45 | 122 |
| voice | 17 | 55 |
| name | 31 | 32 |
| man | 85 | 262 |
| woman | 7 | 4 |
These descriptive counts come from the successful live-source GitHub Actions run. Type-token ratio is sensitive to corpus length, so it should not be treated as a simple measure of one text being “more diverse” than the other.