--- title: "Uruguay — research institutions, scientists and papers" description: "Uruguay: 1 institutions with papers in our library in 1 agglomerations, 3 people with a page, 2 works, 281 citations." canonical: "https://research.elamur.ai/free/en/countries/uruguay/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/countries/uruguay/) [home](https://md.research.elamur.ai/free/en/) / [countries](https://md.research.elamur.ai/free/en/countries/) / Uruguay # Uruguay Country 1 institution in 1 agglomeration · [all countries](https://md.research.elamur.ai/free/en/countries/) [**1** institution with a page](https://md.research.elamur.ai/free/en/countries/uruguay/#institutions) · [**1** agglomeration](https://md.research.elamur.ai/free/en/countries/uruguay/#agglomerations) · [**3** people with a page](https://md.research.elamur.ai/free/en/countries/uruguay/#authors) · [**2** works in our library](https://md.research.elamur.ai/free/en/countries/uruguay/#works) · [**281** citations of their works](https://md.research.elamur.ai/free/en/countries/uruguay/#works) ## About ### Profile Uruguay appears on this site through a single, tightly focused line of research in computational linguistics. All of its work here comes from one university, the [Universidad de la República de Uruguay](https://md.research.elamur.ai/free/en/universities/universidad-de-la-republica-de-uruguay/), located in the capital region of [Montevideo](https://md.research.elamur.ai/free/en/agglomerations/montevideo/). That agglomeration is the only one the country has on this site, and the university is the only institution, so the country's profile here is effectively the profile of one research group's contributions. Those contributions belong to [natural language processing](https://md.research.elamur.ai/free/en/shelves/nlp/) and to [linguistics and languages](https://md.research.elamur.ai/free/en/shelves/linguistics/). They address a specific problem: how to build and evaluate language technology for the Indigenous languages of the Americas, which have very little digital text and are largely missing from the material on which modern multilingual systems are trained. This is a narrow portrait. It reflects only the works and people on this site, not the full range of research carried out in the country. Readers can compare it with other nations through the [list of all countries](https://md.research.elamur.ai/free/en/countries/), and can browse the national [institutions](https://md.research.elamur.ai/free/en/countries/uruguay/institutions/) and [agglomerations](https://md.research.elamur.ai/free/en/countries/uruguay/agglomerations/) directly. ### History On this site, Uruguay's research history is short and recent. It begins in 2021 and continues into 2022. The first work is a report on a shared evaluation campaign. In that campaign, teams built machine translation systems for Indigenous languages of the Americas and compared them under common conditions. A year later came a second work from the same circle of authors. It moved from translation to language understanding. It asked how well large pretrained multilingual models handle reasoning about sentence meaning in languages those models never saw during training. The two works share their central authors, including Arturo Oncevay, Abteen Ebrahimi and Manuel Mager. This continuity suggests a single, connected research program rather than separate efforts. It first established community benchmarks for translation, then extended the same family of languages to a harder test of semantic understanding. No earlier or later works from Uruguayan institutions appear here. ### Research profile The common thread across the country's output on this site is [low-resource languages](https://md.research.elamur.ai/free/en/topics/low-resource-languages/). Every work here falls under that topic, and each treats the scarcity of text and resources as the main obstacle to be measured and overcome. Within that frame, the first strand is [machine translation](https://md.research.elamur.ai/free/en/topics/machine-translation/). The 2021 shared-task report describes an open competition on translating between Spanish and a set of Indigenous languages of the Americas. A shared task of this kind gives many research teams the same training material and the same test sets, so their systems can be compared fairly. The report also helps attract wider attention to languages that mainstream translation research had rarely addressed. According to its summary, the scores achieved were modest, which shows how difficult translation remains when parallel text is so scarce. The second strand joins [language models and pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) with [multilingual and cross-lingual NLP](https://md.research.elamur.ai/free/en/topics/multilingual-nlp/). The 2022 AmericasNLI work starts from a weakness in earlier studies. When researchers tested multilingual models on languages absent from pretraining, they mostly used syntactic tasks. AmericasNLI shifts the test to natural language inference, where a system must decide whether one sentence follows from, contradicts, or is unrelated to another. Because this requires grasping meaning rather than structure alone, it is a more demanding probe of what such models can transfer. The work evaluates zero-shot performance, meaning the model gets no task-specific training data in the target language. It thereby measures how far cross-lingual transfer reaches into what the authors call truly low-resource languages. Both strands also contribute to [evaluation and benchmarks](https://md.research.elamur.ai/free/en/topics/evaluation-benchmarks/). The aim is not only to propose methods but to create shared test resources and protocols that other researchers can reuse. Seen together, the Uruguayan works in our library focus on building measurement tools for underserved languages: a translation benchmark first, then an inference benchmark. ### Leading scientists The first scientist shown for Uruguay is [Luis Chiruzzo](https://md.research.elamur.ai/free/en/authors/luis-chiruzzo/), affiliated with the Universidad de la República de Uruguay. On this site, Chiruzzo's work belongs to the same line of research on language technology for Indigenous and low-resource languages of the Americas. That work connects the university to the international effort behind the AmericasNLP and AmericasNLI resources. Next is [John Ortega](https://md.research.elamur.ai/free/en/authors/john-ortega/), also tied to the Universidad de la República de Uruguay. Ortega's papers here deal with machine translation and evaluation for under-resourced languages. This matches the concerns of the shared task on translating Indigenous languages of the Americas. The third is [Arturo Oncevay](https://md.research.elamur.ai/free/en/authors/arturo-oncevay/), associated with the same university. Oncevay appears among the authors of both of the country's works in our library. That makes Oncevay one of the clearest links between the 2021 translation campaign and the 2022 inference benchmark. Oncevay's research here centers on multilingual methods and on testing how well models handle languages with very little data. The university's own profile also shows [Abteen Ebrahimi](https://md.research.elamur.ai/free/en/authors/abteen-ebrahimi/) and [Angela Fan](https://md.research.elamur.ai/free/en/authors/angela-fan/) among its people. Ebrahimi is first author of AmericasNLI and a co-author of the shared-task report. Fan's presence reflects the collaborative, multi-institution character of the work. ### Landmark works The most cited Uruguayan work in our library is [AmericasNLI: Evaluating Zero-shot Natural Language Understanding of Pretrained Multilingual Models in Truly Low-resource Languages](https://md.research.elamur.ai/free/en/papers/ebrahimi-2022-americasnli-evaluating-zero-shot-natural/), from 2022, led by Abteen Ebrahimi with Manuel Mager, Arturo Oncevay and others. It extends natural language inference to a set of Indigenous languages of the Americas. It then uses the resulting benchmark to examine how pretrained multilingual models behave on languages missing from their training. Its contribution is twofold: a new evaluation resource, and evidence about the limits of zero-shot transfer when a task depends on meaning rather than syntax. Many later papers in our library cite it. The second landmark is [Findings of the AmericasNLP 2021 Shared Task on Open Machine Translation for Indigenous Languages of the Americas](https://md.research.elamur.ai/free/en/papers/mager-2021-findings-americasnlp-2021-shared-task/), from 2021, led by Manuel Mager with Arturo Oncevay, Abteen Ebrahimi and colleagues. It reports the setup, participating systems and results of a community competition on translation into Indigenous languages. It documents both the progress made and how hard the problem remains. As a findings paper, it serves as the reference account of the campaign and as a starting point for later translation work on these languages. It, too, is frequently cited by works here. ### Labs and units The only institution through which Uruguay appears on this site is the Universidad de la República de Uruguay, a university in Montevideo. All of the country's people and works here pass through it. No separate departments, laboratories or research units are named. The university's research on this site sits within natural language processing and linguistics, concentrated on low-resource and Indigenous languages. ### Collaborations The country's works here are collaborative by nature. Shared tasks and benchmarks of this kind usually involve many contributors across institutions and countries, and the author lists reflect that. Researchers connected to the Universidad de la República de Uruguay appear alongside co-authors such as Manuel Mager, Abteen Ebrahimi and Angela Fan. The recurring pairing of Mager, Oncevay and Ebrahimi across the 2021 and 2022 works points to a stable core team. That team organized resources for the Indigenous languages of the Americas and then reused and expanded them. Within Uruguay itself, no links between universities and companies can be traced on this site, since a single university accounts for all national activity here. ### Timeline 2021: The country's first work in our library appears, the findings report of the AmericasNLP shared task on open machine translation for Indigenous languages of the Americas. It sets out common data and evaluation for translation in very low-resource conditions. It also brings together the group of authors who would continue the line of research. 2022: The same circle publishes AmericasNLI, moving from translation to zero-shot natural language understanding. It tests pretrained multilingual models on inference in Indigenous languages absent from their training. It is the most cited Uruguayan work in our library and rounds out the country's current presence here: a compact, coherent program of benchmark-building for languages that language technology has largely overlooked. ## Where ## Agglomerations (1) 1. [**Montevideo**1 institution*281 citations*](https://md.research.elamur.ai/free/en/agglomerations/montevideo/) ## Institutions (1) ### 🎓 Universities (1) **[Universidad de la República de Uruguay](https://md.research.elamur.ai/free/en/universities/universidad-de-la-republica-de-uruguay/)**University · [Montevideo](https://md.research.elamur.ai/free/en/agglomerations/montevideo/), Uruguay*2 works · 281 citations* ## People (3) Authors with a page confirmed at its institutions, the most cited first. [🧑‍🔬**Luis Chiruzzo** Universidad de la República de Uruguay 3 papers here · 486 citations](https://md.research.elamur.ai/free/en/authors/luis-chiruzzo/) · [🧑‍🔬**John Ortega** Universidad de la República de Uruguay 2 papers here · 458 citations](https://md.research.elamur.ai/free/en/authors/john-ortega/) · [🧑‍🔬**Arturo Oncevay** Universidad de la República de Uruguay 3 papers here · 305 citations](https://md.research.elamur.ai/free/en/authors/arturo-oncevay/) ## Works in our library (2) Works of its institutions, the most cited first. 1. [**AmericasNLI: Evaluating Zero-shot Natural Language Understanding of Pretrained Multilingual Models in Truly Low-resource Languages**Abteen Ebrahimi, Manuel Mager, Arturo Oncevay, et al.*2022 · 163 citations*](https://md.research.elamur.ai/free/en/papers/ebrahimi-2022-americasnli-evaluating-zero-shot-natural/) 2. [**Findings of the AmericasNLP 2021 Shared Task on Open Machine Translation for Indigenous Languages of the Americas**Manuel Mager, Arturo Oncevay, Abteen Ebrahimi, et al.*2021 · 118 citations*](https://md.research.elamur.ai/free/en/papers/mager-2021-findings-americasnlp-2021-shared-task/) ## Concepts, shelves and topics Of its works in our library: the number of works on each and its rating among them. 📚 Shelves 2[Natural language processing · 1 work · 12.8](https://md.research.elamur.ai/free/en/shelves/nlp/) · [Linguistics & languages · 1 work · 10.9](https://md.research.elamur.ai/free/en/shelves/linguistics/) 🧭 Topics 5[Low-resource languages · 2 works · 7.9](https://md.research.elamur.ai/free/en/topics/low-resource-languages/) · [Language models & pretraining · 1 work · 4.3](https://md.research.elamur.ai/free/en/topics/language-models/) · [Multilingual & cross-lingual NLP · 1 work · 4.3](https://md.research.elamur.ai/free/en/topics/multilingual-nlp/) · [Evaluation & benchmarks · 1 work · 3.6](https://md.research.elamur.ai/free/en/topics/evaluation-benchmarks/) · [Machine translation · 1 work · 3.6](https://md.research.elamur.ai/free/en/topics/machine-translation/) ## Works by year ### 2020s [**2022** 1 work](https://md.research.elamur.ai/free/en/papers/ebrahimi-2022-americasnli-evaluating-zero-shot-natural/) · [**2021** 1 work](https://md.research.elamur.ai/free/en/papers/mager-2021-findings-americasnlp-2021-shared-task/) - [← back to](https://md.research.elamur.ai/free/en/countries/) — countries