Mariana Romanyshyn

For the past ten years, Mariana Romanyshyn has been leading the team of computational linguists at Grammarly, focusing on the development of error correction and text improvement algorithms. She has extensive experience in developing NLP applications, ranging from foundational solutions like syntactic parsing to features such as sentiment analysis and fact extraction. Mariana is passionate about developing tools and resources for Ukrainian language processing. She serves as the general organizer of the Ukrainian NLP workshop, co-leads the lang-uk initiative, and is a member of the Expert Committee on the Development of AI in Ukraine under the Ministry of Digital Transformation of Ukraine. An educator by calling, Mariana enjoys organizing summer schools and workshops, teaching NLP courses, and supervising diploma papers at the Ukrainian Catholic University. Mariana loves exploring the boundaries of large language models, but her primary interest remains in structural linguistics as a means for formalizing natural language.

From Low-Resource to Sovereign:
The Path of Ukrainian Natural Language Processing

Abstract

Ten years ago, Ukrainian was considered a low-resource language, with no open-source tools or resources available — but that was a decade ago.

In this talk, we will showcase the latest achievements in Ukrainian language processing, while also addressing the unique challenges that this language continues to face. We will observe how shared tasks and workshops, industry and academia, as well as independent communities, are shaping the dynamics of language development and driving innovation in natural language processing (NLP).

We will discuss the limitations of open-source, multilingual large language models (LLMs) in tackling various tasks related to the Ukrainian language and elaborate on the necessity of developing a sovereign LLM specifically designed for Ukrainian.

This talk will be of interest to professionals in academia and industry who want to gain insights into the current state and prospects of Ukrainian NLP, and researchers working on low-resource languages.