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.
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.
| 인물 | 국가 | 이정표 | 나이 / 기록 |
|---|---|---|---|
| Sawsan Ahmed | 미국 | 2021년 12월, 12세로 브로워드 칼리지 최연소 졸업 | 12세 |
| Arvind Mahankali | 미국 | 2013년 13세로 스크립스 내셔널 스펠링 비 우승 | 13세 |
| Kerry Close | 미국 | 2006년 13세로 스크립스 내셔널 스펠링 비 우승 | 13세 |
| Shania Muhammad | 미국 | 15세로 랭스턴 대학교 최연소 졸업 | 15세 |
| Zeyneb N. Kaya | 미국 | 2024년 Regeneron 과학 인재 탐색 대회 5위 | 17세 |
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.
Where is Zeyneb Kaya now?
No source used here reports her university or later work. Nothing after March 2024 is documented on this page.
Is Zeyneb N. Kaya on Wikipedia?
She has no article of her own in the sources used here; her placement is listed on the Society for Science's site.
Quick quiz: how well do you know Zeyneb’s story?
What was Zeyneb N. Kaya's project called?
Why do AI systems struggle with endangered languages?
Where did she place in the 2024 Regeneron Science Talent Search?
Sources
- Regeneron STS 2024 (top 10) — Society for Science
- Regeneron Science Talent Search 2024 Finalists — Society for Science
- Regeneron Science Talent Search 2025 Awards More Than $1.8 Million to High School Seniors — Society for Science, 2025-03-11
Checked by the Geniuses.club editorial team. Ages are given as of September 2026.
