You can follow 95% of Love Island with about 1,250 word families.
I turned all 38 episodes of Love Island USA season 4 into a 224,815-word corpus and measured how much English a viewer needs to follow it. Everyday English needs about 3,000 word families for the same coverage.
Word families needed to understand each share of the show's speech, from V4/tables/vocabulary_coverage.csv. The bars run to the dashed line at 3,000, the published estimate for 95% coverage of everyday spoken English (Adolphs and Schmitt 2003; Nation 2006).
The question
How hard is the show's English?
Reality dating shows feel repetitive. "Feels repetitive" isn't a measurement, so I scraped the subtitle transcripts for every episode of season 4 and measured the language the way a corpus linguist would: how varied it is, how hard it is to read, and how many words you need to know to understand it.
A set of Python scripts counts and lemmatizes the words, fits a Zipf curve, scores readability with textstat, tags parts of speech with NLTK and scores sentiment with TextBlob.
The head of the curve
A handful of words do most of the work
Rank every word by how often it's spoken and the top 100 make up 61% of all the talking. The top three are "you", "to" and "like". Frequency against rank follows Zipf's law closely, with R² = 0.98 on log-log axes.
Reading level
Grade 2.8 by Flesch-Kincaid
Flesch-Kincaid puts the dialogue at grade 2.8, with every episode between 2.3 and 3.7, and a Flesch Reading Ease of 90.2. Formulas that weight long or unfamiliar words read higher: SMOG gives 6.9. 97.9% of the words spoken are on the Dale-Chall list of 3,000 familiar English words.
Across the season
Positive in tone, flat in variety
- Lexical diversity averages an MTLD of 59.96 per episode and shows no trend across the season (Spearman ρ = −0.03, p = 0.87).
- TextBlob sentiment is positive in every episode and highly subjective.
- Adjectives are 8.4% of words; verbs are 24.5% and pronouns 13.6%.
- "Like" appears 28 times per 1,000 words.
Limitations
One season, read from subtitles
This covers one season of one show. Subtitles miss overlapping talk, accents and speed, all of which make speech harder to follow, so the coverage figure is a lower bound on difficulty. Word families are approximated by lemmatization, and the readability formulas disagree with each other on this dialogue, from grade 3 to about 7.
Read the paper (PDF) · Full write-up · Code and corpus on GitHub