PlainSpell data
The corpus in numbers
According to the PlainSpell database, the corpus holds 6,918,744 words and 3,375,996 confusable pairs across five languages, sourced from Wiktionary via kaikki.org, a structured export of over 1,000,000 English dictionary entries (CC BY-SA). Every figure below is counted directly from the live database, see the methodology for how each is derived, and is free to cite or available to download as a CSV dataset.
At a glance
PlainSpell indexes 6,918,744 words across 5 languages, with 3,375,996 confusable pairs, 27,821 homophones and 1,974,632 generated misspelling variants.
- 6.92M
- words indexed
- 3.38M
- confusable pairs
- 27,821
- homophones
- 1.97M
- misspelling variants
Source: Wiktionary (kaikki.org, CC BY-SA) + open word-frequency list. Data vintage May 2026.
Key findings
- PlainSpell catalogues 6,918,744 words across 5 languages (French, German, Spanish, English, Portuguese), all derived from Wiktionary.
- French is the largest of the 5 dictionaries at 4,485,239 words, more than the other four languages combined.
- The corpus maps 3,375,996 confusable word pairs: visually or phonetically similar words a spell-checker cannot disambiguate on its own.
- German has the most confusable pairs (2,006,359), a by-product of its highly productive compound-word formation.
- French leads in catalogued homophones (21,890 words that share a pronunciation), reflecting its many silent letters and merged vowel sounds.
- The corpus also generates 1,974,632 edit-distance misspelling variants for spell-check and typo research.
Each figure is computed live from the PlainSpell database (data vintage May 2026) and free to cite (CC BY-SA).
Words indexed by language
| Language | Words | Confusables | Homophones |
|---|---|---|---|
| 🇫🇷 French | | 440,172 | 21,890 |
| 🇩🇪 German | | 2,006,359 | 2,859 |
| 🇪🇸 Spanish | | 323,831 | 812 |
| 🇺🇸 English | | 530,003 | 2,182 |
| 🇧🇷 Portuguese | | 75,631 | 78 |
| All languages | 6,918,744 | 3,375,996 | 27,821 |
Counts are read live from the PlainSpell database (data vintage May 2026). Misspelling variants are generated by edit-distance from each headword, not observed corpus frequencies.
Download the data (CSV) - per-language counts, CC BY-SA, free to reuse.
Go deeper
From the totals to the individual records.
- Explore any language in full, definitions, IPA, etymology and misspellings. English A–Z
- See the cross-language rankings: hardest to spell, most confusable, largest homophone groups. Rankings
- Read how the corpus is built and refreshed. Methodology