This is the frequency list our free Esperanto course is built on: the 3,000 most frequent Esperanto roots, in order, each with its citation form, its x-system spelling, its part of speech and a short English gloss. We are publishing it as open data because a good, freely reusable Esperanto frequency list was surprisingly hard to find, and the next person building a course, a flashcard deck or a reading tool should not have to start from scratch.
Download
- CSV: esperanto-3000-frequency.csv (about 270 KB, UTF-8, one root per row)
- JSON: esperanto-3000-frequency.json (the same rows, plus the licence and field notes)
The columns are:
| Column | What it holds |
|---|---|
rank | 1 is the most frequent root |
lemma | the citation form, with ĉ ĝ ĥ ĵ ŝ ŭ |
x_system | the same form in the x-system: cx, gx |
part_of_speech | as on the word's page |
gloss_en | a short English gloss; a semicolon separates distinct senses |
url | the root's page on this site, with examples and audio |
Licence and how to cite it
The list is released under Creative Commons Attribution 4.0. Use it for anything, commercial or not, as long as you credit it. A citation that works:
Kilo Lingo (2026). Esperanto 3,000-root frequency list, version 2026-09-25. https://kilolingo.com/resources/esperanto-frequency-list
Part of the corpus behind ranks 1,001 to 3,000 is Tatoeba, whose sentences are licensed CC BY 2.0 FR, so please credit Tatoeba as well.
How it was built
Ranks 1 to 1,000 come from about 2.95 million words of public-domain Esperanto prose from Project Gutenberg, most of it written between 1887 and 1930: Zamenhof and his contemporaries. Dictionaries, grammars and textbooks were excluded, because they list every word once or are mostly English. We built our own count rather than borrow one because the convenient existing lists are share-alike, which would have tied the licence of everything built on them.
Ranks 1,001 to 3,000 come from a blend of three corpora, combined by relative frequency rather than raw counts (otherwise the biggest corpus simply wins): the same public-domain prose (weight 1), the Esperanto sentences on Tatoeba (weight 2, the largest body of modern, everyday Esperanto available) and the Esperanto film subtitles in the OPUS OpenSubtitles collection (weight 1, the most conversational register there is). Each corpus had a small noise floor applied before blending, to drop OCR errors and one-off typos.
From the raw counts to the list, every step was curation rather than arithmetic:
- Endings off, affixes on. Grammatical endings (-o, -a, -e, -j, -n, -as, -is, -os, -us, -u, -i and the participles) are removed by rule. Derivational affixes are not: stripping re- by machine also mangles reĝo, which is not re- plus anything. An affixed form keeps its own entry only when its meaning is not compositional (malsanulejo, a hospital, which any learner can build from sana plus three affixes, is left to the affix track).
- Proper nouns out. Modern corpora inflate names wildly: Tatoeba's stock character Tom ranked 17th in the first blend. Names, cities and countries were removed.
- Junk and archaisms reviewed by hand. OCR debris went, and a handful of genuinely archaic words the old prose over-represents were dropped. Words where the corpus signal was unclear were dropped rather than guessed at.
- One page per root. Each root in the list has a page on this site with examples, so every gloss was written alongside real example sentences rather than copied from a dictionary.
Known biases
A frequency list is only as good as its corpus, so here is what this one leans towards:
- The top thousand is literary and old. Ranks 1 to 1,000 come from prose written a century ago. Function words and core vocabulary are stable across that century, so the head of the list is sound, but religious and epistolary words sit higher than they would in a modern count, and technology is absent from that band.
- The modern blend is sentence-shaped. Tatoeba sentences are short and written to illustrate words, which favours everyday vocabulary and first and second person; subtitles favour dialogue. Both are corrections the old prose needed, and both have their own tilt.
- Ranks are fuzzier further down. Above a few hundred, neighbouring ranks are not meaningfully different. Use the list as an order to learn in, not a measurement.
Using it to learn
The list is the spine of our Esperanto course, taught a hundred roots per chapter alongside the 40 official affixes, with a graded story at every milestone that uses no word the course has not yet taught. If you just want the words, the 1,000 most common Esperanto words and the full list are browsable on the site, with every root linked to its page. For the method behind learning them in this order, see how to learn Esperanto.
If you use the list in something, we would genuinely like to hear about it, and if you find an error, tell us and we will fix it in the next version.