Zeyneb N. Kaya — Biographie de Jeune Prodige

Biographie

The problem MADLIBS addresses

Machine translation runs on parallel text: the same passage in two languages, in enormous quantity. For English and Spanish that material exists in the billions of sentences. For a language with a few thousand living speakers and no printed corpus, it barely exists at all, and the standard methods have nothing to learn from.

That is why translation tools cover a few dozen languages well and thousands badly or not at all. The gap is not a lack of interest among linguists; it is a lack of data, and it falls hardest on exactly the languages closest to disappearing.

What she built

The Society for Science lists Kaya fifth in its 2024 top ten, for MADLIBS, described as an algorithm to help AI translate endangered languages. The Society publishes the placement and the description; the paper itself is not reproduced there, and this page does not summarise a method it has not read.

What the description does establish is the shape of the work: an algorithmic contribution aimed at low-resource translation, rather than a dataset or an app. That is the branch of the field where progress is measured in how little data a system needs, not how much it can consume.

Why endangered languages are a scientific problem

A language carries what its speakers know - plant names, navigation, kinship, law - in a form that does not survive translation intact. When the last fluent speakers die, the loss is not only cultural. It removes a record that no other language holds in the same way.

Technology has cut both ways. Digital life accelerates the shift towards a handful of dominant languages, and it is also the only plausible route to documenting and teaching smaller ones at scale. Work like MADLIBS sits precisely on that line.

The 2024 competition

The Regeneron Science Talent Search names 40 finalists a year from roughly 2,000 entries, and places ten of them after a week of questioning by scientists in Washington. In 2024 first place went to Achyuta Rajaram of Phillips Exeter Academy, with Thomas Cong second, Michelle Wei third, Nathan Wei fourth and Kaya fifth.

The Society's account of the 84th edition, the following year, records $3.1 million awarded in total and $1.8 million shared among finalists. The 2024 pages used here do not state that year's figures, so no prize amount is attributed to Kaya.

LE TEST GRATUIT

Et le vôtre ?

Vingt questions, huit minutes au chronomètre et un centile mesuré face à tous ceux qui l’ont passé. Sans inscription.

Passer le test de QI →

What is documented, and what is not

Documented: fifth place in 2024, the age of 17, Saratoga High School in California, the project name MADLIBS and its purpose.

Not documented in these sources: which languages were tested, what the algorithm does in detail, what results it achieved, any mentor or institution, any publication, and where she went afterwards. Readers who want the method should look for the paper rather than for summaries of it.

Chronologie des Réalisations

Zeyneb N. Kaya — every milestone on this page, placed by age.

2024

Nommée l'une des 40 finalistes duNommée l'une des 40 finalistes du Regeneron Science Talent Search

2024

Classée cinquième à 17 ans pour MADLIBS, un algClassée cinquième à 17 ans pour MADLIBS, un algorithme visant la traduction automatique des langues en danger

2025

La Society for Science enregistre l'édLa Society for Science enregistre l'édition suivante attribuant 3,1 millions de dollars à travers la compétition

Comparaison Historique

Personne	Pays	Étape	Âge / Donnée
Sawsan Ahmed	États-Unis	La plus jeune diplômée du Broward College, à 12 ans, en décembre 2021	12 ans
Arvind Mahankali	États-Unis	A gagné le Scripps National Spelling Bee 2013 à 13 ans	13 ans
Kerry Close	États-Unis	A gagné le Scripps National Spelling Bee 2006 à 13 ans	13 ans
Shania Muhammad	États-Unis	La plus jeune diplômée de Langston University, à 15 ans	15 ans
Zeyneb N. Kaya	États-Unis	Cinquième place, Regeneron Science Talent Search 2024	17 ans

What is documented — and what is not

✓ Documented

She placed fifth in the 2024 Regeneron Science Talent Search
Her project, MADLIBS, aimed to help AI translate endangered languages
She was 17 and attended Saratoga High School, California

? Unverified, disputed or wrong

Which languages the work covered and what results it achieved
The technical detail of the algorithm
Any mentor, laboratory or publication
Where she is studying now

Questions people ask about Zeyneb N. Kaya

How old is Zeyneb N. Kaya?

She was 17 at the 2024 Regeneron Science Talent Search; her birth year is recorded as 2007. No date of birth appears in these sources.

What is MADLIBS?

The name of her project: an algorithm intended to help artificial intelligence translate endangered languages, as described in the Society for Science's 2024 listing.

Where did she place?

Fifth in the 2024 Regeneron Science Talent Search.

Which school did she attend?

Saratoga High School in California.

Which languages did she work on?

No source used here names the languages tested. The description states the purpose of the algorithm, not its test set.

Why is translating endangered languages hard for AI?

Machine translation depends on very large amounts of parallel text. Languages with few speakers and little written material do not supply it, so standard methods have almost nothing to learn from.