Duygu Ataman

Assistant Professor of Computer Science

New York University

I am an assistant professor and faculty fellow at the Department of Computer Science, Courant Institute of Mathematical Sciences, New York University. My research interests include studying generalization in artificial intelligence systems, particularly generative models, through self-supervised and multi-task learning methods. I also actively work on developing resources and methodology for low-resource natural language processing.

The main goal of my research is extending the applicability of language technologies in more languages and tasks. Language technology is a highly promising tool with the potential of transforming how we use and benefit from education and the media, although state-of-the-art approaches are still not competitive enough to be deployed in a large portion of the world languages. This problem is related to the nature of assumptions made in formulating the statistical language models that usually fail to generalize to languages with various syntactic typology, in particular the ones with relatively high data sparsity. I find this highly constrained design problem quite intriguing due to its potential in benefiting a tremendous amount of future applications; at the same time, in line with my background in engineering, where I had specialised on developing optimized software solutions for real-time computation-intensive multimedia technology. Previous to joining the New York University, I was a post-doctoral researcher and lecturer at the University of Zürich, and an applied research scientist intern at the Amazon Alexa research, where I worked on developing novel methods for improving generative models in low-resource and morphologically-rich languages.

Interests

Artificial Intelligence
Multimedia Technology
Computational Linguistics

Education

Ph.D. in Information Engineering and Computer Science, 2019

University of Trento
Ph.D. in Informatics (Visiting Post-graduate Student), 2018

University of Edinburgh
M.Sc. in Embedded Systems and Multimedia Technology, 2015

University of Leuven
B.Sc. in Electrical and Electronics Engineering, 2013

Middle East Technical University

News

Last updated on Apr 25, 2024 1 min read

Applicability of multilingual language models across different settings are conventionally evaluated using task and language-specific performance in downstream benchmarks, which are often lacking in many languages. Our recent study explores a set of efficient and reliable measures that could aid in computing more information related to the generalization capability of language models in cross-lingual zero-shot settings. Preprint available on the arxiv.

Last updated on Apr 25, 2024 1 min read

The 1st annual meeting of ACL SIGTURK will take place in collocation with ACL in Bangkok, Thailand on August 15-16, 2024. More details are available on the workshop website.

Last updated on Feb 17, 2024 1 min read

I will be in Singapore December 6th - 11th to serve as program chair of the Multilingual Representation Learning Workshop at EMNLP 2023.

Last updated on Apr 25, 2024 1 min read

Congratulations to Grace Wang and Tia Chen for getting their paper accepted at the Eval4NLP Workshop at IJCNLP-AACL 2023. Their research project investigates how evaluation metrics for generative models generalize across various forms of rephrasing in languages with distinct typology. Paper accessible through the workshop proceedings.

Last updated on Apr 25, 2024 1 min read

I was selected as one of the principal investigator awardees of the Microsoft Accelerating Foundation Models Research Program. The project will investigate novel methods for improving accessibility to foundation models in low-resourced languages.

Last updated on Sep 29, 2023 1 min read

I will be serving as an Area Chair in the area of “Less-resourced/Endangered/Less-studied Languages” track at LREC-COLING 2024.

See all posts