Where Romanian words come from
In DEX, Romania’s standard dictionary, almost half the words are French or were made in Romanian from French words: between 48.1% and 52.6%, depending on how words with several origins are counted. 13.7% are inherited from Latin. Yet in the dialogue of films and series subtitled in Romanian, almost four words in five are Latin: 79.0%.
Each word takes the colour of the language it comes from. Blue is inherited Latin: words passed from speaker to speaker, from the Romans to today.
Mâine dimineață mergem la piață să cumpărăm pâine, brânză și cafea.
Words inherited from Latin: 8 of 11.
The small words get you first: și (and), de (of), în (in), la (at), să (to), nu (no), eu (I), tu (you), a fi (to be), a avea (to have) are all inherited from Latin. Then try a sentence with no French word: most of the time you manage it at once. And yet the dictionary has 3.8 times as many French words as words inherited from Latin.
Six well-known texts, coloured by the same rules. In “Miorița”, 23 of 27 words are inherited from Latin and none is French; in Article 1 of the Constitution, 20 of 61 words are French.
A fost odată ca-n povești,
A fost ca niciodată,
Din rude mari împărătești,
O prea frumoasă fată.
Și era una la părinți Și mândră-n toate cele,
Cum e Fecioara între sfinți Și luna între stele.
Inherited Latin: 29 of 35 words · French: 2
Words used: 37, distinct: 28. Inherited Latin: 78%, several origins: 5%, two words in one spelling: 0%.
Pe-un picior de plai,
Pe-o gură de rai,
Iată vin în cale,
Se cobor la vale Trei turme de miei,
Cu trei ciobănei.
Unu-i moldovan,
Unu-i ungurean Și unu-i vrâncean.
Inherited Latin: 23 of 27 words · French: 0
Words used: 33, distinct: 24. Inherited Latin: 82%, several origins: 0%, two words in one spelling: 3%.
Stau câteodată și-mi aduc aminte ce vremi și ce oameni mai erau în părțile noastre pe când începusem și eu, drăgăliță-Doamne, a mă ridica băiețaș la casa părinților mei, în satul Humulești, din târg drept peste apa Neamțului; sat mare și vesel, împărțit în trei părți, care se țin tot de una: Vatra satului, Delenii și Bejenii.
Inherited Latin: 43 of 54 words · French: 1
Words used: 57, distinct: 45. Inherited Latin: 79%, several origins: 2%, two words in one spelling: 0%.
Tatăl nostru, Care ești în ceruri, sfințească-se numele Tău; Vie împărăția Ta; facă-se voia Ta, precum în cer și pe pământ. Pâinea noastră cea spre ființă dă-ne-o nouă astăzi; Și ne iartă nouă greșelile noastre, precum și noi iertăm greșiților noștri; Și nu ne duce pe noi în ispită, ci ne izbăvește de cel rău.
Inherited Latin: 48 of 55 words · French: 1
Words used: 59, distinct: 39. Inherited Latin: 85%, several origins: 0%, two words in one spelling: 3%.
România este stat național, suveran și independent, unitar și indivizibil. Forma de guvernământ a statului român este republica. România este stat de drept, democratic și social, în care demnitatea omului, drepturile și libertățile cetățenilor, libera dezvoltare a personalității umane, dreptatea și pluralismul politic reprezintă valori supreme, în spiritul tradițiilor democratice ale poporului român și idealurilor Revoluției din decembrie 1989, și sunt garantate.
Inherited Latin: 30 of 61 words · French: 20
Words used: 61, distinct: 44. Inherited Latin: 46%, several origins: 25%, two words in one spelling: 7%.
Fraților români, Timpul mântuirii noastre a venit; popolul român se deșteaptă la glasul trâmbiței îngerului mântuirii și își cunoaște dreptul său de suveran. Pace vouă, pentru că vi se vestește libertate vouă!
Inherited Latin: 24 of 32 words · French: 2
Words used: 32, distinct: 27. Inherited Latin: 75%, several origins: 3%, two words in one spelling: 0%.
Deșteaptă-te, române, din somnul cel de moarte,
În care te-adânciră barbarii de tirani
Acum ori niciodată croiește-ți altă soarte,
La care să se-nchine și cruzii tăi dușmani.
Acum ori niciodată să dăm dovezi în lume
Că-n aste mâni mai curge un sânge de roman,
Și că-n a noastre piepturi păstrăm cu fală-un nume
Triumfător în lupte, un nume de Traian!
Inherited Latin: 48 of 57 words · French: 3
Words used: 66, distinct: 47. Inherited Latin: 79%, several origins: 8%, two words in one spelling: 3%.
1. consumator - orice persoană fizică sau grup de persoane fizice constituite în asociații, așa cum sunt definite la art. 2 pct. 2 din Ordonanța Guvernului nr. 21/1992 privind protecția consumatorilor, republicată, cu modificările și completările ulterioare;
2. profesionist - orice persoană fizică sau juridică, publică sau privată, care acționează în cadrul activității sale comerciale, industriale sau de producție, artizanale ori liberale în legătură cu contractele care intră sub incidența prezentei ordonanțe de urgență, precum și orice persoană care acționează în același scop, în numele sau pe seama acesteia;
Inherited Latin: 34 of 79 words · French: 29
Words used: 79, distinct: 56. Inherited Latin: 43%, several origins: 23%, two words in one spelling: 4%.
Type any Romanian word: see what DEX says about its origin, what Wiktionary says, how often it is used today and how often it appeared in prose a century or two ago.
The search covers the 59,075 DEX ’98 entries that have an origin, plus the frequent words DEX does not explain separately.
Some Latin words reached Romanian by two roads. Once by word of mouth, changing like everything spoken: directus became drept (straight, right). Again, centuries later, from books or through French, barely changed: direct. Sometimes the meanings drifted far apart: tonus gave both tun (cannon) and ton (tone). DEX ’98 has 39 such pairs in which both words are still in use; all of them are here.
Ten everyday Romanian words. Pick the language each comes from, according to DEX ’98. Some will surprise you.
The usual count assigns one source language to each word. Here every origin named in the note is kept. When the DEX note names several languages (grav: “Din fr. grave, lat. gravis”), the word is assigned to the class several origins. When the same spelling can be two words with different origins (masă “table”, from Latin, and masă “mass”, from French), it is counted in the class two words, one spelling, and the page does not choose between the two.
| Class | DEX ’98 entries (59,075) | Film dialogue, words as used | Wikipedia, words as used |
|---|---|---|---|
| Inherited Latin | 12.9% | 78.4% | 61.4% |
| Learned Latin | 1.7% | 0.5% | 2.3% |
| French | 41.5% | 4.5% | 11.2% |
| Italian | 1.2% | 0.3% | 0.4% |
| Old Slavic | 5.0% | 2.7% | 2.3% |
| Neighbouring Slavic languages | 3.9% | 1.2% | 0.7% |
| Hungarian | 1.2% | 0.4% | 0.8% |
| Turkish | 2.2% | 0.1% | 0.4% |
| Greek | 1.5% | 0.5% | 0.5% |
| German | 1.9% | 0.1% | 0.3% |
| English | 1.2% | 0.2% | 0.2% |
| Other languages | 0.2% | — | — |
| Unknown; DEX compares it first with Albanian | 0.5% | 0.3% | 0.2% |
| Unknown origin | 6.6% | 2.1% | 1.5% |
| Made in Romanian | 1.8% | 0.5% | 0.4% |
| Several origins | 16.4% | 6.6% | 15.0% |
| Two words, one spelling | 0.4% | 1.6% | 2.5% |
In film dialogue, 78.4% of words as used are inherited Latin, 6.6% have several origins in their note and 1.6% are spellings that can be two words. In the dictionary French stays the largest layer (41.5%), and 16.4% of entries have several origins.
The count in the first screen gives 79.0% for inherited Latin in film dialogue: it assigns every word with several origins under the first language of its note and every ambiguous spelling under its commonest homonym. The difference from 78.4% is 0.6 percentage points.
Dimitrie Macrea counted every poem Eminescu published in his lifetime: 3,607 distinct words, used 33,846 times in all. Inherited Latin, with its Romanian derivatives, was 48.68% of the distinct words and 83% of the words as used. In Candrea’s dictionary (43,269 words) the same layer was 20.58%. The figures are from Dacoromania X (1938–1941), pp. 364, 370–371 and 373.
A separate count covers 94 poems from Wikisource’s list of poems published in Eminescu’s lifetime: 34,579 words as used and 3,216 distinct words, classified with DEX ’98 and the rules on this page.
| Class | Macrea (Candrea, 1931) | Here (DEX ’98) |
|---|---|---|
| Inherited Latin, with its derivatives, of distinct words | 48.68% | 47.6% |
| Inherited Latin, with its derivatives, of words as used | 83% | 81.1% |
| Old and Church Slavic, of words as used | 6.18% | 4.3% |
| French, of words as used | 2.52% | 1.9% |
| Learned Latin, of words as used | 1.13% | 0.6% |
| Unknown origin, of words as used | 1.80% | 3.1% |
| Several origins | — | 3.7% |
| Two words, one spelling | — | 1.8% |
The figures are not identical and cannot be: a different text (Wikisource’s spelling), a different dictionary, different rules for variants. Macrea counted variants as separate words, which is why he has 3,607 distinct words and this page 3,216. The share of inherited Latin differs by at most two percentage points: 1.1 for distinct words and 1.9 for words as used. This count also separates two classes the dictionary count does not show: 3.7% of the words as used have several origins in their note, and 1.8% are spellings that can be two words.
You can read that Romanian is two-thirds French, that it is three-quarters Latin, or that French gave it three words in five. Each number has a source. What differs is the thing being counted. Below are the same words, counted five ways.
59,075 words with an origin; those made in Romanian form their own layer.
36,119 words: this is where “two-thirds French” comes from.
The page’s main count: abandonare (abandonment) is credited to French, acasă (at home) to Latin.
332 million words of subtitles, each counted every time it occurs; shares are of the 98.8% of words that were recognised.
17.2 million words from 41,357 articles.
In a dictionary each word counts once: masă (table) as much as multilateralism. In speech a word counts every time it is said. Half of all the Latin in film dialogue comes from just 16 words, among them fi, avea, nu, el, să, de, eu, vrea. French needs 332 words to reach half of its occurrences.
Prepositions, conjunctions, pronouns, articles and the verbs a fi, a avea, a vrea (to be, to have, to want) make up 28.7% of dialogue. Take them out and inherited Latin still has 71.5%, French 12.5%.
DEX ’98 gives an origin for 59,075 entries. Almost two in five (38.9%) were made in Romanian from other Romanian words, as abandonare (abandoning) was from abandona. Leave them aside and 36,119 words from other languages remain, of which French has 66.9%: hence “two-thirds”. Credit them to the language of the word they were made from and French has 52.6%.
Wiktionary barely marks words made in Romanian (2.4%), which is why its categories put French at 59.9% of 47,657 words. Where DEX and Wiktionary classify the same word (32,730 words), they give the same language 92.0% of the time. The different numbers come mainly from what is counted; for 8.0% of the shared words the two sources give different languages.
DEX separates inherited words from those taken from written Latin: it writes “Lat. mensa” for masă and “Din lat.” (from Latin) for a loan such as abrevia. Put the two together and Latin reaches 7.9% of words from other languages; the inherited part is 4.1%. The dictionary does not always keep its own rule: for 85 inherited words, among them om (man), it still writes “Din lat.”. Here they are counted as inherited, following Wiktionary.
For 5,076 of the 26,717 French words, DEX names several languages at once: grav is “from French grave, Latin gravis”. Linguists call this multiple etymology. The main count uses the first language in the note; split these words equally among the languages given and French has 48.1% instead of 52.6%, learned Latin 6.8%.
Masă, the table, comes from Latin; masă, the mass, from French. Da, yes, comes from Slavic; a da, to give, from Latin. The forms such pairs share make up 7.0% of dialogue. They are assigned by a rule described under the method, and the look-up shows both entries.
The dictionary also records names of people, places and brands that became ordinary words.
For each language: how many DEX ’98 entries come directly from it, what share of the dictionary it holds when words made in Romanian are credited to the language they were made from, and what share of the words spoken and written today.
| Language of origin | DEX ’98 entries taken directly | Of the dictionary | Of what is said | Of Wikipedia |
|---|---|---|---|---|
| French | 24,158 | 52.6% | 8.9% | 22.9% |
| Inherited Latin | 1,496 | 13.7% | 79.0% | 62.0% |
| Old Slavic | 1,006 | 5.5% | 2.9% | 2.5% |
| Neighbouring Slavic languages | 1,241 | 4.3% | 1.6% | 1.0% |
| Learned Latin | 1,354 | 3.4% | 2.0% | 5.6% |
| German | 1,092 | 2.5% | 0.2% | 0.6% |
| Turkish | 940 | 2.4% | 0.4% | 0.3% |
| Greek | 629 | 1.9% | 0.6% | 0.8% |
| Italian | 768 | 1.7% | 0.6% | 1.1% |
| English | 812 | 1.5% | 0.3% | 0.3% |
| Hungarian | 348 | 1.3% | 0.6% | 0.8% |
| Other languages | 89 | 0.2% | 0.0% | 0.1% |
| Made in Romanian | 22,956 | 1.8% | 0.5% | 0.4% |
| Unknown origin | 2,186 | 7.3% | 2.5% | 1.8% |
It depends what you count. Of DEX ’98 entries, 13.7% are inherited from Latin or made in Romanian from inherited words, and 3.4% were taken from written Latin. Of the words spoken in film dialogue, 79.0% are inherited Latin.
In DEX ’98, between 48.1% and 52.6% of entries are French or made from French words, depending on how words with several origins are counted. Of words that came directly from other languages, French has 66.9%. In speech, only 8.9% of spoken words are French.
It is true of dictionary words that came directly from other languages: French has 66.9% of them. It is not true of spoken words: in film dialogue, 79.0% is inherited Latin and 8.9% is French.
Each language of origin has a different share in film dialogue and in Wikipedia. French words are 2.6 times more frequent in Wikipedia than in film dialogue, learned Latin 2.9 times. Speech gives more room to inherited Latin, to the neighbouring Slavic languages (1.6 times, mostly through da, yes) and to Turkish (1.3 times, with haide, rahat, cafea: come on, damn, coffee).
The Proclamation of Islaz and the constitutions that followed are publicly available, dated political and constitutional texts covering a century and a half. In 1848 French had 7.5% of the words; in the 2003 Constitution, 29.8%. Inherited Latin fell from 78.1% to 55.3%. Turkish words are almost entirely absent from all nine texts.
| Year | Text | Words | Inherited Latin | French | Learned Latin | Slavic |
|---|---|---|---|---|---|---|
| 1848 | Proclamația de la Islaz | 4,238 | 78.1% | 7.5% | 2.9% | 5.1% |
| 1866 | Constituția României | 4,838 | 71.0% | 16.3% | 5.9% | 1.9% |
| 1923 | Constituția României | 6,994 | 66.1% | 19.2% | 7.8% | 2.2% |
| 1938 | Constituția României | 5,654 | 64.4% | 19.9% | 8.4% | 2.1% |
| 1948 | Constituția Republicii Populare Române | 4,010 | 58.3% | 22.7% | 9.9% | 2.4% |
| 1952 | Constituția Republicii Populare Române | 5,321 | 52.7% | 26.4% | 7.3% | 4.7% |
| 1965 | Constituția Republicii Socialiste România | 6,805 | 53.0% | 26.3% | 9.7% | 3.3% |
| 1991 | Constituția României | 9,457 | 55.2% | 29.8% | 7.6% | 1.9% |
| 2003 | Constituția României, revizuită | 11,567 | 55.3% | 29.8% | 7.8% | 1.6% |
In the dated prose of Romanian authors on Wikisource, French rises from 3.9% (1825–1849) to 8.7% (1925–1949). Inherited Latin stays the largest, from 80.8% to 77.3%. The series covers 1825–1949, the years for which Wikisource has enough dated prose.
| Period | Texts | Authors | Words | Inherited Latin | French |
|---|---|---|---|---|---|
| 1825–1849 | 450 | 18 | 406,207 | 80.8% | 3.9% |
| 1850–1874 | 953 | 41 | 1,957,434 | 79.6% | 5.1% |
| 1875–1899 | 1,935 | 54 | 3,089,249 | 79.2% | 6.0% |
| 1900–1924 | 2,321 | 86 | 2,848,168 | 76.5% | 8.0% |
| 1925–1949 | 967 | 60 | 3,468,369 | 77.3% | 8.7% |
For each: what it means, its share and which words carry it in speech. Select a word to see it in the look-up.
Words handed down from spoken Latin, generation after generation, changing shape on the way: aqua became apă (water), oculus became ochi (eye). DEX marks them “Lat.”, without “Din” (from).
1,496 entries in DEX ’98 · 13.7% of the dictionary · 79.0% of what is said · 62.0% of Wikipedia
Words taken straight from the Latin of books, barely changed: abrevia (to abbreviate), abecedar (primer), regulă (rule). DEX marks them “Din lat.”, like any other loan.
1,354 entries in DEX ’98 · 3.4% of the dictionary · 2.0% of what is said · 5.6% of Wikipedia
Most spoken: regulă, ultim, familie, caz, oră, moment, spera, persoană, doctor, ocupa
The largest layer of the dictionary. For many of these words DEX also gives a second source, usually Latin: grav (serious) is “from French grave, Latin gravis”.
24,158 entries in DEX ’98 · 52.6% of the dictionary · 8.9% of what is said · 22.9% of Wikipedia
Most spoken: problemă, tip, termina, exista, mașină, serios, idee, deja, dată, deveni
Loans from Old Slavic, including Church Slavonic, used in worship and administration. DEX writes all of them “Din sl.” and does not separate the spoken route from the written one.
1,006 entries in DEX ’98 · 5.5% of the dictionary · 2.9% of what is said · 2.5% of Wikipedia
Most spoken: trebui, nevoie, prea, prieten, iubi, opri, treabă, omorî, drum, primi
Loans from the modern Slavic languages, which DEX names one by one: Bulgarian, Serbo-Croatian, Ukrainian, Russian, Polish.
1,241 entries in DEX ’98 · 4.3% of the dictionary · 1.6% of what is said · 1.0% of Wikipedia
Most spoken: da, grijă, poliție, ciudat, grozav, școală, ba, păstra, grăbi, obicei
Romanian words from Bulgarian, Serbian, Russian, Ukrainian and Polish
Words DEX gives as “from Hungarian”: oraș (town), gând (thought), fel (kind), seamă (account).
348 entries in DEX ’98 · 1.3% of the dictionary · 0.6% of what is said · 0.8% of Wikipedia
Most spoken: fel, hei, oraș, seamă, gând, locui, pahar, bănui, uriaș, marfă
DEX separates modern Greek (“ngr.”) from ancient Greek (“gr.”); here they are counted together. Sigur (sure), măcar (at least) and stradă (street) are given as modern Greek.
629 entries in DEX ’98 · 1.9% of the dictionary · 0.6% of what is said · 0.8% of Wikipedia
Most spoken: sigur, măcar, stradă, rest, ah, lipsi, pat, sosi, scop, englez
Words DEX gives as “from Turkish”: haide (come on), cafea (coffee), dușman (enemy), cutie (box). The few from Tatar are counted here too.
940 entries in DEX ’98 · 2.4% of the dictionary · 0.4% of what is said · 0.3% of Wikipedia
Most spoken: haide, rahat, haideți, cafea, cutie, șerif, hain, dușman, murdar, habar
Words DEX gives as “from German”, including the Transylvanian Saxon dialect: spital (hospital), polițist (policeman), pantof (shoe).
1,092 entries in DEX ’98 · 2.5% of the dictionary · 0.2% of what is said · 0.6% of Wikipedia
Most spoken: polițist, spital, referi, pantof, laborator, proiect, firmă, lecție, vest, țigară
Words DEX gives as “from Italian”: cameră (room), scuză (excuse), căpitan (captain).
768 entries in DEX ’98 · 1.7% of the dictionary · 0.6% of what is said · 1.1% of Wikipedia
Most spoken: scuza, cameră, reuși, căpitan, oferi, distruge, stat, proces, incredibil, speranță
Words DEX gives as “from English”: meci (match), lider (leader), weekend. The dictionary dates from 1998, so newer loans are missing.
812 entries in DEX ’98 · 1.5% of the dictionary · 0.3% of what is said · 0.3% of Wikipedia
Most spoken: dolar, cip, york, meci, weekend, computer, lord, lider, fotbal, team
Spanish, the Romani language, Arabic, Hebrew, Portuguese and others, 89 entries together; 25 of them are given by DEX as Romani, among them mișto (cool) and baftă (luck).
89 entries in DEX ’98 · 0.2% of the dictionary · 0.0% of what is said · 0.1% of Wikipedia
Most spoken: mișto, rom, nasol, baftă, gagică, marijuana, country, călău, rodeo, danci
Romanian words from other languages: Romani, Spanish, Arabic, Hebrew
Words Romanian built from its own words: derivatives (abandonare, abandoning, from abandona), compounds (acasă, at home, from a and casă). 8,130 of them are participles and verbal nouns that DEX lists as separate entries. In the main count each is credited to the language of the words it was made from.
22,956 entries in DEX ’98 · 1.8% of the dictionary · 0.5% of what is said · 0.4% of Wikipedia
Most spoken: pentru, din, dacă, ceva, vorbi, cel, despre, mulțumi, lucru, sau
DEX writes “Et. nec.” (etymology unknown) or only offers a comparison (“Cf.”). For 84 of them the comparison is with Albanian, as for brad (fir), barză (stork) and buză (lip); hence the hypothesis that they come from the language spoken here before Latin.
2,186 entries in DEX ’98 · 7.3% of the dictionary · 2.5% of what is said · 1.8% of Wikipedia
Most spoken: dar, găsi, copil, puțin, iar, ban, naibă, băiat, gata, adică
All 59,075 DEX ’98 entries with an origin, in one file: the word, the dictionary’s etymology note, the origin assigned here, the language a Romanian-made word is credited to, Wiktionary’s verdict, and how often it occurs in film dialogue and in Wikipedia. Licence GNU GPL v2 or later, like the dexonline database it comes from.
The labels come from a program that reads the etymology notes of DEX ’98. To see how often it is wrong, 327 notes were drawn at random (seed 20261010) from all layers, with more from the rare ones. The notes were read separately by two readers, Claude Code and Codex, two language models, who were given the same written rules (below). Codex did not see the program’s answers. Claude Code knew the program and its rules, and had seen its answer for six of the notes, printed when the sample was drawn. The sample was read only by these two models, not by a person.
The program gave the same origins as both readers on 318 of 327 notes (97.2%, 95% confidence interval 94.9%–98.5%). On the 321 notes whose answer had not been seen: 312 (97.2%, 94.8%–98.5%). The two readers matched each other on 324 notes. On six notes both agreed against the program: a note that compares the word with Albanian and Bulgarian without naming a source, which the program classifies under Albanian comparisons, two “Din lat.” words the program classifies as inherited Latin on Wiktionary’s evidence, a note that points at another homonym (“Cf. copil.”), a base word with no note, and an origin written in parentheses, which the program ignores. The last is the only outright mistake.
All 327 notes, with the program’s answer and each reader’s: error-sample.csv. The readers were given the rules below, so the sample checks the reading of the notes against those rules, not the correctness of the etymologies in DEX ’98.
Each word’s origin comes from the note at the end of its definitions in the 1998 edition of the Explanatory Dictionary of the Romanian Language (DEX ’98), as it appears in the public dexonline database of 3 September 2026. Of the dictionaries in the public copy of the database, DEX ’98 is the one that gives an etymology for the whole vocabulary; DEX ’96, which gives one only for part of it, fills in where DEX ’98 has none. Each homonym keeps its own origin (masă¹, masă²). The note is read in the order it is written: in “from French grave, Latin gravis” French counts.
Words made in Romanian (“V. abandona”, “Din abur”, “A + casă”) are credited, in equal parts, to the language of the words they were made from, following the chain back to a word from another language. Where the note names no base word, the word stays “made in Romanian”.
A second verdict comes from Wiktionary, for 32,730 words. It decides only when DEX writes “Din lat.” for a word Wiktionary gives as inherited and, for homonyms of different origin, which of them is the everyday word.
Speech is the frequency list of Romanian subtitles in OpenSubtitles 2018: 332 million words of film and television dialogue. Writing is Romanian Wikipedia: 41,357 articles, 17.2 million words. Older prose is dated text from Wikisource, with at most two million characters from any one author in each quarter-century.
From the word in a text to the word in the dictionary. Each form (mergem, casei) is linked to its headword through dexonline’s 2.27 million inflected forms. A form written with diacritics is matched exactly. One written without them (in, cand) is shared among its possible spellings according to how often each occurs in Wikipedia. A form belonging to several words is shared according to what each collects from its unambiguous forms; for a few dozen very frequent forms the decision follows grammar: e is a form of a fi (to be), not the letter, and și on its own is the conjunction. Clusters such as s-a or n-am are split into their words.
Limits. DEX ’98 is the 1998 edition; the revisions of 2009, 2012 and 2016 are not in the public database, and newer English loans are missing. A dictionary marks a loan more readily than an inheritance, and “unknown origin” is the dictionary’s verdict, not linguistics’ final word. Subtitles are dialogue written for the screen, often translated, not spontaneous speech. In dialogue, 7.0% of words sit on forms of homonyms with different origins, and their result depends on the rule above.
This page counts what Romanian took from other languages; the reverse, 392 Romanian words taken by 13 languages, each read in the dictionary of the language that received it, is counted in Romanian words in other languages.
More on the site: How many words does Romanian have? · The word cemetery (in Romanian) · The Dacian words · A day in words (in Romanian) · What the words meant · Borrowed clothes