Elamur Research⌕
☰
home / papers / 2026 / A Corpus-Based Profiling of Regional English Variants in Global Media: Insights from Olympic Journalism

A Corpus-Based Profiling of Regional English Variants in Global Media: Insights from Olympic Journalism

Summary

The study asks whether English-language news from outlets in different countries shows measurable stylistic signatures. The author treats these as institutional editorial styles, not as fixed sociolinguistic varieties. To hold the topic roughly constant, the author builds the Olympic Journalism English Media Corpus: about 1,810 Olympics-related articles (~1.0 million tokens, 2020–2023) from five English-language outlets in the US, China, Spain, Costa Rica and Mexico. The last three are pooled into one Spanish-speaking-context class. Texts are processed with Stanford CoreNLP, and 164 theory-driven features covering lexis, syntax, readability, sentiment and dependency structure are extracted. SVM classifiers are trained on these features, on GPT embeddings, and on both together. The most discriminative features are then kept using SVM importance and Bonferroni-corrected ANOVA. The best model reaches an F1 of 97.2. Explicit features alone still achieve a macro F1 of 74.9, which points to stylistic rather than purely topical signals. The resulting profiles characterize American coverage as informational and neutral, Chinese coverage as formal and analytical, and Spanish-context coverage as expressive. The approach offers a reusable template for computational World Englishes research.

Key points

  • Corpus composition:

    • The New York Times: 780 articles, 568,478 tokens
    • China Daily: 704 articles, 305,437 tokens
    • Sur in English: 153 articles, 63,253 tokens
    • The Tico Times: 59 articles, 39,068 tokens
    • Mexico News Daily: 114 articles, 51,198 tokens
    • The corpus spans Tokyo 2020, Beijing 2022, and coverage during the surrounding Olympic cycle.

  • Features: There are 164 features in total, split into 67 basic ones (POS distributions, TO/RP tags, function/content ratios, TTR) and 97 advanced ones. The advanced set includes:

    • lexical richness measures
    • Activity/Descriptivity indices
    • nominality
    • readability scores (Flesch–Kincaid, SMOG, LIX)
    • sentiment metrics
    • mean dependency distance

  • Classification setups: Four SVM configurations were compared: basic features, advanced features, GPT embeddings (text-embedding-3-large, reported as 4096 dimensions), and embeddings combined with advanced features. All use an 80/20 stratified split.

  • Results and feature analysis: The optimized F1 is 97.2, and the feature-only macro F1 is 74.9. Features with SVM importance below 0.005 were discarded, and ANOVA (p < .05, Bonferroni) was applied across the three style groups. Verb ratio, nominality and readability emerged as key differentiators.

  • Limitations: The authors acknowledge several:

    • The corpus covers a single genre.
    • Each region is represented by few outlets.
    • Journalists' L1 backgrounds are unknown.
    • Topical leakage from differing national sports focus is possible.
    • Only one embedding/classifier pairing was tested.

    Metadata and preprocessing scripts are promised for release.

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 10.63317/3kzf8hoic5ht

@inproceedings{mao2026corpus,
  title = {{A Corpus-Based Profiling of Regional English Variants in Global Media: Insights from Olympic Journalism}},
  author = {Mao, Felix},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/3kzf8hoic5ht},
  url = {https://doi.org/10.63317/3kzf8hoic5ht}
}

References

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

  1. Lucie Flek. 2020. Returning the n to nlp: Towards contextually personalized classification models. In Proceedings of the 58th annual meeting of the association for computational linguistics, pages 7828–7838. find
  2. Edward Gibson. 2000. The dependency locality theory: A distance-based theory of linguistic complexity. Image, language, brain. ↗
  3. Heike Hänlein. 1999. Studies in authorship recognition: a corpus based approach. Journal of Computational Linguistics, 13(3):221–235. find
  4. Francis Heylighen and Jean-Marc Dewaele. 1999. Formality of language: definition, measurement and behavioral determinants. Interner Bericht, Center “Leo Apostel”, Vrije Universiteit Brüssel 4.1. find
  5. Braj B Kachru. 1990. World englishes and applied linguistics. World Englishes, 9(1):3–20. ↗
  6. Roby Marlina and Zhichang Xu. 2018. English as a lingua franca. The TESOL encyclopedia of English language teaching, pages 1–13. ↗
  7. Tom McArthur. 2001. World english and world englishes: Trends, tensions, varieties, and standards. Language teaching, 34(1):1–20. ↗
  8. Xiaxing Pan, Hui Qiu, and Haitao Liu. 2015. Golden section in chinese contemporary poetry. Glottometrics, 32:55–62. find
  9. Alastair Pennycook. 2009. English and globalization. In The Routledge companion to English language studies, pages 125–133. Routledge. ↗
  10. Brent W Ritchie, Richard Shipway, and P Monica Chien. 2010. The role of the media in influencing residents’ support for the 2012 olympic games. International Journal of Event and Festival Management, 1(3):202–219. Min Song and Tamy Chambers. Text mining with the stanford corenlp. Measuring scholarly impact: Methods and practice, pages 215–234. ↗
  11. Benedikt Szmrecsanyi and Bernd Kortmann. 2009. The morphosyntax of varieties of english worldwide: A quantitative perspective. Lingua, 119(11):1643–1663. ↗
  12. Maite Taboada. 2016. Sentiment analysis: An overview from linguistics. Annual Review of Linguistics, 2.1. ↗
  13. Zhen Tao, Dinghao Xi, Zhiyu Li, Liumin Tang, and Wei Xu. 2024. Cat-llm: Prompting large language models with text style definition for chinese article-style transfer. arXiv preprint arXiv:2401.05707. ↗
  14. R.W. Van Hout and A. R. Vermeer. 2007. Comparing measures of lexical richness. Cambridge University Press. ↗
  15. Haiyan Zhou, Yue Jiang, and Letao Wang. 2023. Are daojing and dejing stylistically independent of each other: A stylometric analysis with activity and descriptivity. Digital Scholarship in the Humanities, 38(1):434–450. find
  16. Sanja Štajner and et al. 2012. What can readability measures really tell us about text complexity. Proceedings of workshop on natural language processing for improving textual accessibility. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/mao-2026-corpus-based-profiling-regional-english/