Language library
Real languages, answered through the same questions you answer. Nobody can
change these — take a copy and the copy is yours.
Every number elsewhere in ProtoLingua is a share of real languages. These
are 15 of them, from
11 families,
so the comparison can be with a language rather than with a percentage.
English
Indo-European
Analytic, rigidly SVO, with prepositions and almost no case left. Its two surviving cases are on the pronouns, and word order does the work case used to do.
296 answers ·
36 from a database ·
260 from a grammar ·
67% match
Frenchfrançais
Indo-European
Romance, SVO, with prepositions, two genders and no case on nouns. Most adjectives follow the noun, negation is a discontinuous pair, and the definite article is obligatory where English drops it.
300 answers ·
30 from a database ·
270 from a grammar ·
61% match
Modern Greekελληνικά
Indo-European
Fusional, three genders, four cases, and no dominant word order: the case endings free the constituents, so order carries information structure rather than grammatical relations. Subject pronouns are dropped because the verb already carries the person.
299 answers ·
32 from a database ·
267 from a grammar ·
33% match
Russianрусский язык
Indo-European
Six cases, three genders, aspect before tense, no articles at all, and a present-tense copula that is simply absent. Word order is free because the case endings say who did what.
300 answers ·
34 from a database ·
266 from a grammar ·
26% match
Finnishsuomi
Uralic
Agglutinative, fifteen cases, no gender and no articles. Negation is an inflecting auxiliary rather than a particle, and vowel harmony decides the shape of nearly every suffix.
300 answers ·
34 from a database ·
266 from a grammar ·
19% match
Basqueeuskara
Isolate
A language isolate, verb-final, and ergative-absolutive: the subject of a transitive verb takes a case the subject of an intransitive one does not. The auxiliary agrees with up to three arguments at once.
303 answers ·
28 from a database ·
275 from a grammar ·
28% match
Georgianქართული
Kartvelian
Kartvelian, verb-final, with a case system whose alignment changes with the tense series and a verb that agrees with both subject and object. The consonant inventory has ejectives and long onset clusters.
296 answers ·
31 from a database ·
265 from a grammar ·
39% match
TurkishTürkçe
Turkic
Agglutinating and verb-final, with a long chain of suffixes on the verb, front-back and rounding vowel harmony, and an evidential distinction built into the past tense.
301 answers ·
36 from a database ·
265 from a grammar ·
23% match
Hindiहिन्दी
Indo-European
Verb-final, with postpositions and a case system that becomes ergative in the perfective: the agent of a completed transitive action takes a marker the same agent does not take otherwise. Two genders, and a four-way stop contrast.
302 answers ·
32 from a database ·
270 from a grammar ·
43% match
Egyptian Arabicمصري
Afro-Asiatic
Semitic. Most of the morphology is a consonantal root interleaved with a vowel pattern rather than a chain of affixes, so the app's affix questions describe it only in part. Adjectives and demonstratives follow the noun; the definite article is a prefix.
300 answers ·
29 from a database ·
271 from a grammar ·
47% match
SwahiliKiswahili
Niger-Congo
SVO like English and nothing else like it: a dozen or more noun classes agreeing across the whole clause, prefixes rather than suffixes, and adjectives, numerals and demonstratives all following the noun.
298 answers ·
36 from a database ·
262 from a grammar ·
35% match
Mandarin普通话
Sino-Tibetan
Isolating and tonal. Nothing inflects: tense, aspect and number are separate morphemes or absent, a numeral cannot touch a noun without a classifier between them, and the relative clause precedes its noun while the verb still precedes its object.
297 answers ·
30 from a database ·
267 from a grammar ·
62% match
Vietnamesetiếng Việt
Austroasiatic
Isolating and tonal, with six tones and no inflection at all. Modifiers follow the noun, classifiers are obligatory with numerals, and person is expressed by kinship terms rather than pronouns.
297 answers ·
27 from a database ·
270 from a grammar ·
69% match
Korean한국어
Koreanic
Verb-final and agglutinating, with case particles on the noun and a speech-level suffix on every finite verb. There are no articles and no agreement; the relative clause and every modifier precede the noun.
299 answers ·
33 from a database ·
266 from a grammar ·
45% match
Japanese日本語
Japanese
Verb-final and consistently head-final, with postpositions, prenominal relative clauses and a topic particle doing work that English gives to word order.
299 answers ·
37 from a database ·
262 from a grammar ·
54% match
What the two counts mean
From a database means the answer is what WALS, Grambank or
PHOIBLE records for that language. Every one of these is checked: the answer
is fed back through ProtoLingua's own atlas machinery and the result compared
with the published coding, so an answer that does not reproduce its own
source stops the build.
From a grammar means there is no published coding for that
question and the answer was written from reference grammars, which each
entry names. It is a judgement, and it is marked wherever it appears.
Most answers are the second kind, and that is the honest position rather
than a gap: 249 of the 326 questions here have no coding in any
cross-linguistic database, because they ask about script, culture,
conversation and vocabulary rather than about syntax.
What the match figure means
Each entry carries a hand-written translation of the seven test sentences.
Match is how much of that translation ProtoLingua actually
produces when it renders the entry, counted word part by word part, in
order.
It measures ProtoLingua, not the language. The databases
record where a marker sits and whether it exists, never what it is, and the
app carries one suffix for each category. So a language that packs three
categories into one ending loses marks on every inflected word, and an
isolating language scores higher for the same reason. The figure rises when
the app gets better at saying things, and the entries stay as they are.
The translation uses the same two dozen words the entry does, so a low
figure never means the entry picked a different word for "grass". What is
being compared is word order, agreement, case, tense and the marker forms.
What these are not
They are not complete descriptions. A library language answers this app's
questions, and this app's questions are a particular cut through a grammar:
there is no morphophonology here, no prosody beyond stress, and the
vocabulary is a token two dozen words rather than a dictionary. For the real
thing, each entry names the reference grammars it was written from.