Guest Lecture Friday May 29th, 2026: Liisa Rätsep

Professional/academic contact person: Nikita Shvetsov

Target group: Students and staff at Department of Computer Science.

Welcome!

-----------

Date: Friday May 29th, 2026, 10:15-12:00
Place: Teknologibygget 1.022
Title: Neural Speech Synthesis for Low-Resource Finno-Ugric Languages
Lecturer: Liisa Rätsep
Streaming URL: https://uit.cloud.panopto.eu/Panopto/Pages/Embed.aspx?id=26843c77-af3f-4ad7-a825-b4510083c6ad

Abstract

In this talk, we present our ongoing work on developing neural text-to-speech models for Finno-Ugric languages. These include higher-resource languages such as Estonian, as well as many endangered and under-resourced languages and dialects, including several Sámi languages. We discuss the challenges involved in working with low-resource and morphologically complex languages, including the limited availability of high-quality and diverse speech data, complex text normalization pipelines, and ethical and legal considerations, and present some of the technical solutions and methodological approaches we have explored to address them.

Bio

Liisa Rätsep is an Estonian computer scientist and language technology researcher who currently serves as the Head of Applied Natural Language Processing (NLP) at the University of Tartu, Estonia. Her work focuses primarily on advanced speech technology, including natural-sounding, expressive neural speech synthesis (text-to-speech) for the Estonian language as well as smaller, lower-resource Finno-Ugric languages.

Kortnytt fra Institutt for informatikk

Rishaug, Tina tina.rishaug@uit.no Forsker/rådgiver