HiTZketan: Deep bilingual speech to speech translation

Ibon Saratxaga, Inmaculada Hernáez, Eneko Agirre, Aitor Soroa, Eva Navas, German Rigau, Jon Sanchez, Rodrigo Agerri
2023-tik 2024 arte
Finantzaketa egin duen erakundea:
Zenbatekoa guztira:

This proposal aims to develop a speech-to-speech (S2S) translation system for Basque and Spanish, i.e. a system that receives a speech signal in one of the two languages, translates it to the other, and delivers this translation using speech with a personalised voice imitating the original speaker.

The project is a joint proposal of the two research groups of Hitz Center (Ixa and Aholab) and it is intended to boost their collaboration to generate a tool that requires the knowledge from the expertise fields of both of the groups. Both IXA, in the field of NLP and Aholab in speech processing have plenty of experience and resources. The S2S translation system that will be produced will enable Basque language support for state-of-the-art technologies in automatic speech recognition, machine translation and personalised speech synthesis.

This proposal responds to the Global Challenges of the Horizon Europe Pillar II. Specifically speech and natural language processing is aligned with Cluster 4: Digital, Industry & Space in the area of Artificial Intelligence-Data-Robotics.