Photo of Xabier Sarasola

Artificial intelligence & robotics

Xabier Sarasola

Speech synthesis researcher in Elhuyar Foundation

Year Honored
2022

Organization
Elhuyar Foundation

Region
Europe

Xabier Sarasola is a speech synthesis researcher in Elhuyar Foundation and he has worked in multiple text-to-speech projects related with low resource languages, gender equality and cross-lingual synthesis. First, he has developed a multi-speaker speech synthesis system for Basque using neural networks. This system allows a low resource language like Basque to imitate any voice using only a small set of utterances of that voice. He has also used cross-lingual techniques to make the Basque voices speak in other languages that the original speaker does not speak. Combined with speech recognition, these cross-lingual techniques have the potential to create language dubbing systems that maintain the voice characteristics of the original voice. Moreover, he has participated in the development of the speech synthesis of Languedocien Occitan and Gascon Occitan. Finally, following the UNESCO recommendations to overcome gender bias in virtual assistants, he has created a gender ambiguous voice for the adaptation of the virtual assistant Mycroft to the Basque language. These projects have helped to build high quality text-to-speech services for low resource languages, an important step in an increasingly interconnected and digital world.