Inguna Skadina: Machine Translation for less resourced languages: Challenge or opportunity?

The opportunity to use translation tools to overcome language barriers has attracted researchers even before the first computers became available. Although a great deal of effort has been put into the research and development of machine translation (MT) solutions, this technology has also received a lot of critique for translation quality. In 2012 the META-NET Language White Papers showed that only three of the official EU languages can benefit from moderate to good support by machine translation technologies, while there is weak or no support for most of the EU languages. Less resourced morphologically rich languages with rather free word order have been recognized as difficult for automated translation tools. This invited talk will discuss issues related to the creation of MT solutions for morphological rich languages and share experiences in the development of MT solutions for the languages of the three Baltic countries – typical representatives of difficult use case for MT. Different machine translation system architectures will be analysed. The most recent outcomes from the H2020 QT21 project, CEF.AT tenders, and WMT 2017 and WMT 2018 news translation shared tasks will be discussed.

Inguna Skadiņa is a professor at the Faculty of Computing of the University of Latvia (UL) and chief scientific officer at Tilde. She has been working for over 30 years on language technologies for less resourced languages. Her current research interests include language resources and tools, machine translation, and human-computer interaction. I.Skadiņa is a national coordinator of CLARIN research infrastructure in Latvia. She is also senior researcher at the Institute of Mathematics and Computer Science, UL. I.Skadiņa has led and participated in many EU and nationally funded projects related to language technologies, including the scientific coordination of the FP7 ACCURAT project and the ICT PSP META-NORD project, and recently participated in the H2020 QT21 project. Currently she is responsible for the scientific coordination of the large scale national project “Neural Network Modelling for Inflected Natural Languages”. I. Skadiņa is the author of more than 50 publications. She is a member of several professional organisations, expert of the Latvian Council of Sciences.