---
title: "Lexical variation in English language podcasts, editorial media,"
description: "He compares them on simple, inspectable counts: pronouns, part-of-speech shares from NLTK tagging, sentence tense… 21 references."
authors: ["Jussi Karlgren"]
year: 2022
venue: "Northern European Journal of Language Technology"
source: "https://nejlt.ep.liu.se/article/view/3566"
doi: "10.3384/nejlt.2000-1533.2022.3566"
licence: "CC BY 4.0"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/karlgren-2022-lexical-variation-english-language-podcasts/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/karlgren-2022-lexical-variation-english-language-podcasts/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2022](https://md.research.elamur.ai/free/en/papers/2022/page/7/) / Lexical variation in English language podcasts, editorial media, and social media

# Lexical variation in English language podcasts, editorial media, and social media

**Authors:** [Jussi Karlgren](https://md.research.elamur.ai/free/en/authors/jussi-karlgren/) · **Year:** 2022 · **Venue:** [Northern European Journal of Language Technology](https://md.research.elamur.ai/free/en/venues/northern-european-journal-of-language-technology/) ([Northern European Journal of Language Technology 2022](https://md.research.elamur.ai/free/en/venues/northern-european-journal-of-language-technology/2022/)) · **Publisher:** [Northern European Association for Language Technology](https://md.research.elamur.ai/free/en/publishers/northern-european-association-for-language-technology/) · **Licence:** [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://doi.org/10.3384/nejlt.2000-1533.2022.3566) · [PDF](https://nejlt.ep.liu.se/article/download/3566/3547) · [DOI 10.3384/nejlt.2000-1533.2022.3566](https://doi.org/10.3384/nejlt.2000-1533.2022.3566) · [OpenAlex](https://openalex.org/W4292370972)

**Citations:** 4 citations (OpenAlex, 2026-10-14)

## Summary
The paper asks how transcribed podcast speech differs lexically from the kinds of English that NLP systems usually learn from. Karlgren draws 100,000 sentences each from seven collections: AP newswire from TIPSTER, the Blog Authorship Corpus, Twitter posts from 2017, the Spotify podcast transcripts, dialogue from the Film Corpus, a Switchboard subset, and TED talk transcripts. He compares them on simple, inspectable counts: pronouns, part-of-speech shares from NLTK tagging, sentence tense, interjections, negations, amplifiers and attitude words. Values are averaged over repeated resamplings.

Compared with newswire, podcasts are mostly present tense and much richer in pronouns, interjections and negation. They share these traits with social media. They differ from blogs in attitude, with more amplifiers and less negativity.

The author draws two lessons for computational work. First, a model's training-data origin is a real variable and should be documented on the model. Second, noun-centred filtering and evaluation lose more information in conversational, pronoun-heavy genres.

## Key points
- **Data:** seven sources with 100,000 sentences each, split into sentences with NLTK at ".", "!", "?" and paragraph breaks. Figures are averages over resamples, rounded to two significant figures.
- **Tense:** sentences that combined several tenses or contained compound verb sequences were left out.
  - Podcasts have about 41,000 present-tense sentences against 14,000 past-tense.
  - Newswire is roughly reversed, with about 14,000 present against 54,000 past.
- **Pronouns:**
  - First-person singular makes up 4.5% of podcast tokens but only 0.37% of newswire tokens.
  - Second person is 3.0% in podcasts and 0.11% in newswire.
- **Proper nouns:** they account for 12% of newswire, 25% of tweets and only 3.7% of podcasts. Nouns overall drop from 22% in newswire to 14% in podcasts.
- **Gender and source caveats:**
  - Every collection shows a large gap between "he" and "she" forms.
  - The Twitter sample turned out to contain many headlines and press-release posts, despite being chosen to represent real-time talk.

## Interesting ideas
- Profiling a corpus with a small set of transparent counts (pronoun person, tense, proper-noun share, interjections) characterises its genre cheaply and readably, without opaque embeddings.
- Drawing equal-sized sentence samples per source and averaging over repeated draws keeps cross-corpus comparisons stable despite very different corpus sizes.
- Discarding sentences that blend tenses or contain compound verb sequences, instead of forcing a label, keeps a tense statistic clean and honest.
- Stoplists and noun-based referential gold standards remove more meaning from pronoun- and verb-heavy text than from news.
- A corpus chosen to represent one register can turn out to hold another (tweets full of headlines), so a source's assumed genre should be checked against its measured profile.

## For Mova

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

- **I1** (learn): Run such a profile on every training or evaluation corpus so that parser and tagger errors can be explained by register shift rather than guessed at.
- **I2** (learn): Use the same fixed-size, resampled protocol when comparing witness-parser agreement across treebanks or genres, so that differences are not artefacts of sample size.
- **I3** (world): When ordering events, mark mixed-tense clauses as unresolved rather than assigning a default timeline position.
- **I4** (frames): Never strip pronouns or light verbs before frame building, and weight coreference evaluation more heavily on conversational texts such as fables and dialogues.
- **I5** (tree): Before using a treebank for UD dialect conversion or silver data, verify its genre profile so that a "spoken" or "informal" resource really behaves that way.

*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

Also on the byline: [Spotify (Sweden)](https://ror.org/00hbd6420).

📚 Shelves (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 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 · [🖱️ Human-Computer Interaction](https://md.research.elamur.ai/free/en/shelves/computer-science/human-computer-interaction/) · 141 works

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

## Cite

Cite as: Jussi Karlgren. 2022. Lexical variation in English language podcasts, editorial media, and social media. Northern European Journal of Language Technology.

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

```bibtex
@article{karlgren2022lexical,
  title = {{Lexical variation in English language podcasts, editorial media, and social media}},
  author = {Jussi Karlgren},
  journal = {Northern European Journal of Language Technology},
  year = {2022},
  doi = {10.3384/nejlt.2000-1533.2022.3566},
  url = {https://doi.org/10.3384/nejlt.2000-1533.2022.3566}
}
```

- [← previous Joint Coreference Resolution for Zeros and non-Zeros in Arabic](https://md.research.elamur.ai/free/en/papers/aloraini-2022-joint-coreference-resolution-zeros-non/)
- [next → Machines in the media: semantic change in the lexicon of mechanization in 19th-century British newspapers](https://md.research.elamur.ai/free/en/papers/pedrazzini-2022-machines-media-semantic-change-lexicon/)
- [↑ back to papers of 2022](https://md.research.elamur.ai/free/en/papers/2022/page/7/)

## References

21 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Banik, Rounak. 2017. TED Talks: Data about TED Talks. Dataset on Kaggle, Version 3. [find](https://scholar.google.com/scholar?q=Banik%2C+Rounak%2E+2017%2E+TED+Talks%3A+Data+about+TED+Talks%2E+Dataset+on+Kaggle%2C+Version+3%2E)
2. Biber, Douglas. 1991. Variation across speech and writing. Cambridge University Press, Cambridge. [find](https://scholar.google.com/scholar?q=Biber%2C+Douglas%2E+1991%2E+Variation+across+speech+and+writing%2E+Cambridge+University+Press%2C+Cambridge%2E)
3. Biber, Douglas. 1995. Dimensions of register variation: Godfrey, John J and Edward Holliman. 1997. Switchboard-1 Release 2. Linguistic Data Consortium, Philadelphia, 926. [find](https://scholar.google.com/scholar?q=Biber%2C+Douglas%2E+1995%2E+Dimensions+of+register+variation%3A+Godfrey%2C+John+J+and+Edward+Holliman%2E+1997%2E+Switchboard%2D1+Release+2%2E+Linguistic+Data+Consortium%2C+Philadelphia%2C+926%2E)
4. Godfrey, John J, Edward C Holliman, and Jane Mc-Daniel. 1992. SWITCHBOARD: Telephone speech corpus for research and development. In Proceedings of IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), volume 1, pages 517– 520. IEEE. [find](https://scholar.google.com/scholar?q=Godfrey%2C+John+J%2C+Edward+C+Holliman%2C+and+Jane+Mc%2DDaniel%2E+1992%2E+SWITCHBOARD%3A+Telephone+speech+corpus+for+research+and+development%2E+In+Proceedings+of+IEEE+International+Conference+on+Acoustics%2C+Speech%2C+a)
5. Harman, Donna and Mark Liberman. 1993. TIPSTER Complete LDC93T3A Web Download. Linguistic Data Consortium, Philadelphia. [find](https://scholar.google.com/scholar?q=Harman%2C+Donna+and+Mark+Liberman%2E+1993%2E+TIPSTER+Complete+LDC93T3A+Web+Download%2E+Linguistic+Data+Consortium%2C+Philadelphia%2E)
6. Hu, Minqing and Bing Liu. 2004. Mining and summarizing customer reviews. In Proceedings of the 10th ACM SIGKDD International conference on Knowledge discovery and data mining (KDD), pages 168–177. Association for Computing Machinery. [find](https://scholar.google.com/scholar?q=Hu%2C+Minqing+and+Bing+Liu%2E+2004%2E+Mining+and+summarizing+customer+reviews%2E+In+Proceedings+of+the+10th+ACM+SIGKDD+International+conference+on+Knowledge+discovery+and+data+mining+%28KDD%29%2C+pages+168%E2%80%93177%2E+Ass)
7. Jurafsky, Dan, Elizabeth Shriberg, and Debra Biasca. 1997. Switchboard SWBD-DAMSL labeling project coder’s manual, Draft 13. Boulder Institute of Cognitive Science Technical Report. [find](https://scholar.google.com/scholar?q=Jurafsky%2C+Dan%2C+Elizabeth+Shriberg%2C+and+Debra+Biasca%2E+1997%2E+Switchboard+SWBD%2DDAMSL+labeling+project+coder%E2%80%99s+manual%2C+Draft+13%2E+Boulder+Institute+of+Cognitive+Science+Technical+Report%2E)
8. Karlgren, Jussi. 2009. Affect, appeal, and sentiment as factors influencing interaction with multimedia information. In Theseus ImageCLEF workshop on visual information retrieval evaluation, pages 8–11. [find](https://scholar.google.com/scholar?q=Karlgren%2C+Jussi%2E+2009%2E+Affect%2C+appeal%2C+and+sentiment+as+factors+influencing+interaction+with+multimedia+information%2E+In+Theseus+ImageCLEF+workshop+on+visual+information+retrieval+evaluation%2C+pages+8%E2%80%931)
9. Karlgren, Jussi. 2019. How lexical gold standards have effects on the usefulness of text analysis tools for digital scholarship. In International Conference of the Cross-Language Evaluation Forum for European Languages (CLEF), pages 178–184. The CLEF Initiative. [find](https://scholar.google.com/scholar?q=Karlgren%2C+Jussi%2E+2019%2E+How+lexical+gold+standards+have+effects+on+the+usefulness+of+text+analysis+tools+for+digital+scholarship%2E+In+International+Conference+of+the+Cross%2DLanguage+Evaluation+Forum+for+)
10. Karlgren, Jussi, Gunnar Eriksson, Stefano Mizzaro, Paul Clough, Mark Sanderson, Kristofer Franzén, and Preben Hansen. 2004. Reading Between the Lines: Attitudinal expressions in text. In Proceedings of the AAAI Spring Symposium Workshop on Exploring Attitude and Affect in Text: Theories and Applications. American Association for Artificial Intelligence. [find](https://scholar.google.com/scholar?q=Karlgren%2C+Jussi%2C+Gunnar+Eriksson%2C+Stefano+Mizzaro%2C+Paul+Clough%2C+Mark+Sanderson%2C+Kristofer+Franze%CC%81n%2C+and+Preben+Hansen%2E+2004%2E+Reading+Between+the+Lines%3A+Attitudinal+expressions+in+text%2E+In+Proceedings+)
11. Kiritchenko, Svetlana, Xiaodan Zhu, and Saif M Mohammad. 2014. Sentiment analysis of short informal texts. Journal of Artificial Intelligence Research, pages 723–762. [↗](https://doi.org/10.1613/jair.4272)
12. Reitan, Johan, Jørgen Faret, Björn Gambäck, and Lars Bungum. 2015. Negation scope detection for twitter sentiment analysis. In Proceedings of the 6th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis, pages 99– 108. [find](https://scholar.google.com/scholar?q=Reitan%2C+Johan%2C+J%C3%B8rgen+Faret%2C+Bjo%CC%88rn+Gamba%CC%88ck%2C+and+Lars+Bungum%2E+2015%2E+Negation+scope+detection+for+twitter+sentiment+analysis%2E+In+Proceedings+of+the+6th+Workshop+on+Computational+Approaches+to+Subjecti)
13. Schler, Jonathan, Moshe Koppel, Shlomo Argamon, and James W Pennebaker. 2006. Effects of age and gender on blogging. In Proceedings of the AAAI Spring Symposium Workshop on Computational approaches to analyzing weblogs, pages 199–205. American Association for Artificial Intelligence. [find](https://scholar.google.com/scholar?q=Schler%2C+Jonathan%2C+Moshe+Koppel%2C+Shlomo+Argamon%2C+and+James+W+Pennebaker%2E+2006%2E+Effects+of+age+and+gender+on+blogging%2E+In+Proceedings+of+the+AAAI+Spring+Symposium+Workshop+on+Computational+approaches+to)
14. Tanushi, Hideyuki, Hercules Dalianis, Martin Duneld, Maria Kvist, Maria Skeppstedt, and Sumithra Velupillai. 2013. Negation scope delimitation in clinical text using three approaches: NegEx, PyCon-TextNLP, and SynNeg. In 19th Nordic Conference of Computational Linguistics (NODALIDA), pages 387– 474. Linköping University Electronic Press. [find](https://scholar.google.com/scholar?q=Tanushi%2C+Hideyuki%2C+Hercules+Dalianis%2C+Martin+Duneld%2C+Maria+Kvist%2C+Maria+Skeppstedt%2C+and+Sumithra+Velupillai%2E+2013%2E+Negation+scope+delimitation+in+clinical+text+using+three+approaches%3A+NegEx%2C+PyCon%2DTex)
15. Von Klopp, Ana. 1993. Negation: Implications for theories of natural language. Ph.D. thesis, University of Edinburgh. [find](https://scholar.google.com/scholar?q=Von+Klopp%2C+Ana%2E+1993%2E+Negation%3A+Implications+for+theories+of+natural+language%2E+Ph%2ED%2E+thesis%2C+University+of+Edinburgh%2E)
16. Walker, Marilyn A, Grace I Lin, and Jennifer Sawyer. 2012. An Annotated Corpus of Film Dialogue for Learning and Characterizing Character Style. In Proceedings of the 8th International Conference on Language Resources and Evaluation (LREC), pages 1373–1378. European Language Resources Association (ELRA). [find](https://scholar.google.com/scholar?q=Walker%2C+Marilyn+A%2C+Grace+I+Lin%2C+and+Jennifer+Sawyer%2E+2012%2E+An+Annotated+Corpus+of+Film+Dialogue+for+Learning+and+Characterizing+Character+Style%2E+In+Proceedings+of+the+8th+International+Conference+on+L)
17. Mohammad, Saif M, Svetlana Kiritchenko, and Xiaodan Zhu. 2013. [NRC-Canada: Building the state-of-the-art in sentiment analysis of tweets](https://md.research.elamur.ai/free/en/papers/mohammad-2013-nrc-canada-building-state-art/). In Second Joint Conference on Lexical and Computational Semantics (* SEM), pages 321–327. [↗](https://doi.org/10.48550/arxiv.1308.6242)
18. Ong, Walter J. 1982. Orality and literacy. Routledge. [↗](https://doi.org/10.4324/9780203328064)
19. Qu, Yan, James Shanahan, and Janyce Wiebe. 2004. Proceedings of the AAAI Spring Symposium Workshop on Exploring Attitude and Affect in Text: Theories and Applications. American Association for Artificial Intelligence. [find](https://scholar.google.com/scholar?q=Qu%2C+Yan%2C+James+Shanahan%2C+and+Janyce+Wiebe%2E+2004%2E+Proceedings+of+the+AAAI+Spring+Symposium+Workshop+on+Exploring+Attitude+and+Affect+in+Text%3A+Theories+and+Applications%2E+American+Association+for+Artific)
20. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik. 1985. A comprehensive grammar of the English language. Longman, London. [find](https://scholar.google.com/scholar?q=Quirk%2C+Randolph%2C+Sidney+Greenbaum%2C+Geoffrey+Leech%2C+and+Jan+Svartvik%2E+1985%2E+A+comprehensive+grammar+of+the+English+language%2E+Longman%2C+London%2E)
21. Ravi, Kumar and Vadlamani Ravi. 2015. A survey on opinion mining and sentiment analysis: tasks, approaches and applications. Knowledge-Based Systems, 89:14–46. A gradation amplifiers very awfully completely enormously entirely exceedingly excessively extremely fucking fuckin greatly highly hugely immensely intensely particularly radically significantly strongly substantially totally utterly vastly Amplifiers affirmation amplifiers absolutely definitely famously genuinely immaculately overly perfectly really severely surely thoroughly truly undoubtedly B surprise amplifiers amazingly dramatically drastically emphatically exceptionally extraordinarily fantastically horribly incredibly insanely phenomenally remarkably ridiculously strikingly surprisingly terribly unusually wildly wonderfully amazing dramatic drastic emphatic exceptional extraordinary fantastic incredible phenomenal remarkable striking surprising unusual Negations Negations are taken from Quirk et al. (1985, §10.54ff) and Biber (1995). analytic no, not n’t, ain’t, aren’t, arent, aren’t, can’t, cannot, cant, couldn’t,, didn’t, doesn’t, don’t, hadn’t, hasn’t, haven’t, contractions isn’t, mightn’t, mustn’t, shouldn’t, wasn’t,, weren’t, won’t, wouldn’t constructional neither, never, nor, none, nobody, no-one, without, sans, w/o [↗](https://doi.org/10.1016/j.knosys.2015.06.015)

*Updated 1 November 2026*