Each seed holds ten DEX words. The commonest sit at the centre; at the rim are the words no subtitle uses. Tap a seed to read them.One seed, ten DEX words. Tap one.
Vocabulary test · Romanian
How many Romanian words do you know: a test on all 64,907 words of the DEX
For anyone who reads Romanian, native speaker or learner. Five screens show 125 words: 100 drawn at random from the DEX, the Romanian Academy’s standard dictionary, and 25 invented, so guessing does not help. Tap the ones you could explain. In about two minutes you learn how many of its 64,907 words you know, with a margin of error.
The test
Five screens of 25 words#
Tap a word if you could tell someone what it means. If it only looks familiar, leave it.
Your result
How many DEX words you know
The flower
Almost a third of the DEX never occurs in 332 million words of subtitles#
The flower has one seed for every ten of the 64,907 one-word headwords of the DEX. The order comes from Romanian subtitles of films and series: 332.6 million words, counted with all their forms (copii and copilului count towards copil, child). At the centre sit de (of), nu (no), să (the subjunctive marker, “to”), a fi (to be) and și (and). Towards the rim the words thin out.
The grey ring at the rim holds 20,825 words, 32% of the dictionary, that no subtitle uses in any form: caier (a tuft of wool or flax for spinning), codruț (a little forest), zâmbire (the act of smiling), capiște (a pagan temple). Another 13,654 words occur fewer than ten times. Altogether, more than half of the DEX (53%) occurs fewer than ten times in 332.6 million words.
The rest of the dictionary carries almost all of speech. The 4,000 commonest words cover 96.1% of the DEX words heard in subtitles, and the first 8,000 cover 98.5%. So people differ little in the middle of the flower, which most people know; they differ towards the rim. The test draws more words from there.
Writers
How far into the dictionary Romanian writers go#
The same dictionary, read through the words of one text. In Amintiri din copilărie (Memories of My Boyhood), Ion Creangă uses 2,644 DEX words. 178 of them (6.7%) never occur in a subtitle: nandralău, năboi, harchină, jnăpăi, cioșmoli. The poems of Mihai Eminescu’s 1884 volume run to almost the same length (26,652 words against 26,635). Of the 2,678 DEX words he uses, only 39 (1.5%) come from the rim.
Liviu Rebreanu’s novel Ion is six times as long and reaches 6,352 DEX words, 89 of them from the rim. The Constitution has none: 1,262 words, all from the used part of the flower. Only words with an entry in the DEX are counted. Creangă’s regional words that the DEX lacks are left out, so for him the figure is, if anything, too low.
Pick a text
Method
How the test reaches a number#
The dictionary’s words are split into ten steps by how often they occur in subtitles. Eight run from the commonest to those that occur once or twice. The last two hold the words that never occur, split by whether they are in DOOM 3, the 2021 orthographic dictionary. The first two screens take two words from each step. The next three put more words where your answers leave the most doubt, usually the steps where you ticked about half.
On each step what counts is the share of words you ticked, after guessing is taken away. If you ticked 2 of 25 invented words, the test assumes you ticked unknown real words just as often, and subtracts them, with a small correction even for someone who ticked none. A rarer step cannot come out better known than a commoner one; where it does, the two are merged. The final number is the sum over the steps: the known share of each, times the number of words in it.
The interval comes from the same answers, redrawn 400 times at random. The test was tried on 400 simulated readers, each with a vocabulary fixed in advance and one of five guessing habits, from none to 15% of unknown words. Its typical error was about 2,700 words (root mean square error), and the interval held the true number in 90% of cases. Someone who ticks many invented words comes out with too high a number and too narrow an interval. Nothing here is a ranking: the answers are sent nowhere, so there is no average of the people who took it.
The test measures recognition: whether you know what a word means when you see it, not whether you use it. It misses words that entered Romanian after the 2009 DEX, slang, proper names and multi-word expressions.
Questions#
How many words does Romanian have?
It depends what you count. The 2009 DEX has 64,907 headwords written as one word, without a hyphen, of at least two letters, with homonyms counted once. DOOM 3, the orthographic standard, covers about 65,000; the Academy’s thesaurus dictionary reaches 125,000 words, 175,000 with variants. How many words does Romanian have? explains each figure.
How many words does a person know?
For Romanian, no measurement on a large sample of speakers was found. For English, Marc Brysbaert and colleagues estimated from online tests that an average 20-year-old American knows 42,000 lemmas, from 27,000 for the lowest 5% to 52,000 for the highest 5%, and adds about 6,000 between 20 and 60, one every two days. Their figures cannot be set directly against these: the word list is different, and so is the question people were asked.
I learnt Romanian at home, abroad. Is the test for me?
Yes. Heritage speakers often know the words of home and family well and the words of school and the press less. The flower shows where your words sit, and the list after your result links some of the words you missed to their definitions in dexonline.
Why are there invented words?
Without them, anyone who ticks out of politeness or haste would come out with a huge vocabulary. The invented words were made by swapping one syllable of a real word and checked against every form in dexonline, the official Scrabble list and the subtitles. Anything that resembled a regional word, a technical term, slang or a typo was then removed. Whatever share of them you tick is subtracted.
What does it mean to “know” a word?
In the test, that you could explain it to someone, even roughly. A word that only looks familiar is not ticked. People who tick more strictly come out with a lower number, and the test cannot tell. So the interval says only how sure the number is for the way you answered.
Why are some new words missing?
The list is that of the 2009 DEX, the latest complete edition in the public dexonline database. Words added to dictionaries since then are not in the test, so the result counts DEX words only.
Is my result sent anywhere?
No. Your answers are sent nowhere. The result is kept on this device only, so you can find it again, until you press “Clear the result”. A link you send carries only the result’s numbers, not the words you ticked.
Sources#
- dexonline, public database dump (lexemes, inflected forms, DEX ’09 headwords); dexonline licence page
- Academia Română, Institutul de Lingvistică „Iorgu Iordan – Al. Rosetti”, DEX, 2nd ed. revised (Univers Enciclopedic Gold, 2009)
- Hermit Dave, FrequencyWords: Romanian word counts from OpenSubtitles 2018 (CC BY-SA 4.0)
- Lison, P., Tiedemann, J. & Kouylekov, M. (2018). OpenSubtitles2018: Statistical rescoring of sentence alignments in large, noisy parallel corpora. LREC 2018
- Brysbaert, M., Stevens, M., Mandera, P. & Keuleers, E. (2016). How many words do we know? Frontiers in Psychology 7:1116
- Lemhöfer, K. & Broersma, M. (2012). Introducing LexTALE: a quick and valid lexical test. Behavior Research Methods 44
- Keuleers, E. & Brysbaert, M. (2010). Wuggy: a multilingual pseudoword generator. Behavior Research Methods 42
- van Heuven, W. J. B. et al. (2014). SUBTLEX-UK and the Zipf scale. Quarterly Journal of Experimental Psychology 67
- Wikisource (Romanian): Eminescu, Poesii (1884); Creangă, Amintiri din copilărie; Caragiale, O scrisoare pierdută; Slavici, Moara cu noroc; Rebreanu, Ion; Constituția României (2003)