MARI Grand Seminar – Large Language Models and Low Resource Languages

Watch our two-hour grand seminar on Large Language Models and Low Resource Languages. The event included a keynote by Dr. Monojit Choudhury titled “Predicting, Explaining and Optimizing Performance of LLMs across Languages,” where he discussed whether massively multilingual language models (MMLM) can be leveraged to predict the accuracy of cross-lingual zero-shot and few-shot transfer for a task on target languages with little or no test data. He also gave an overview of Project LITMUS – Linguistically Informed Training and Testing of Multilingual Systems, which involved building several ML models for performance prediction and discuss the what was learnt about the factors that influence cross-lingual transfer.

The talk was followed by a panel discussion with experts from academia and research; including Dr. Monojit Chowdhury, Dr. Edward Ombui, Dr. Sunayana Sitaram, Dr. David Adelani, and moderated by Maxamed Axmed.

Predicting, Explaining and Optimizing Performance of LLMs across Languages

Given a massively multilingual language models (MMLM), can we predict the accuracy of cross-lingual zero-shot and few-shot transfer for a task on target languages with little or no test data? This seemingly impossible task, if solved, can have several potential benefits. First, we could estimate the performance of a model even in languages where a test set is not available, and/or building one is difficult. Second, one can predict training data configurations that would give certain desired performance across a set of languages, and accordingly strategize data collection plans; this in turn can lead to linguistically fair MMLM-based models. Third, as a byproduct, we would know which factors influence cross-lingual transfer. In this talk, I will give an overview of Project LITMUS – Linguistically Informed Training and Testing of Multilingual Systems, where we build several ML models for performance prediction; besides their applications, I will discuss what we learn about the factors that influence cross-lingual transfer.

Speaker Details

Keynote Speaker

Dr. Monojit Choudhury is a Principal Data and Applied Scientist in Turing, India. Where his team builds large universal language models that forms the backbone of various Microsoft products. Prior to this, he was a Principal researcher at Microsoft Research Lab India, and still strongly collaborates with his colleagues from MSR. His research interests cut across the areas of Linguistics, Cognition, Computation and Society. He has a B.Tech and PhD in Computer Science and Engineering from IIT Kharagpur, and had been at Microsoft Research since 2007. 

Panelists

Dr. Edward Ombui (opens in new tab) is the Dean of the School of Science and Technology at Africa Nazarene University. He earned his PhD in Computer Science from the University of Nairobi. Dr. Ombui is a highly accomplished educator with over 15 years of university administration, teaching, and research experience. He is widely recognized for his contributions to the field of Natural Language Processing (NLP) for low resource languages, with cutting-edge research and innovation in this area. Besides, Dr. Ombui’s active research in the field of NLP with a particular focus on machine learning and deep learning techniques, has seen him receive numerous grants and awards, and publish in top-tier journals and international conferences organized by the Association for Computational Linguistics, the Institute of Electrical and Electronics Engineers, among others.

Dr. Sunayana Sitaram is a Senior Researcher at Microsoft Research India. Her research goal is to enable inclusive universal empowerment through technology. Her current area of research is on making Large Language Models be useful to everyone on the planet by improving their multilingual performance and infusing local culture and values. Sunayana also serves as the director of the MSR India Research Fellow program, which exposes bright young researchers to a world-class research environment and prepares them for careers in research, engineering and entrepreneurship. Prior to joining MSRI as a Post Doc Researcher, Sunayana completed her MS and PhD at the Language Technologies Institute, Carnegie Mellon University in 2015. 

Dr. David Adelani (opens in new tab) is a Research Fellow or DeepMind Academic Fellow in the department of computer Science at University College London, and a member of the leadership of Masakhane – a grassroots organization whose mission is to strengthen and spur natural language processing (NLP) research in African languages, for Africans, by Africans. He was formerly a PhD student of computer science at the department of language science and technology in Saarland University. His research focuses on NLP for under-resourced languages, especially African languages, multilingual representation learning, machine translation, and privacy in NLP.

Date:
Speakers:
Monojit Choudhury, Edward Ombui, Sunayana Sitaram, David Adelani
Affiliation:
Microsoft Research, Africa Nazarene University, Microsoft Research India, University College London

Watch Next