Marta Ruiz Costa-Jussà, awarded with a Starting Grant from the ERC

The researcher and professor at the UPC, Marta Ruiz Costa-Jussà, has received a Starting Grant awarded by the European Research Council (ERC). This European grant for research of excellence will contribute to the development of a project that explores new methods of machine translation of text and voice.

 

Marta Ruiz Costa-Jussà, a researcher at the department of Computer Science, has recently won a grant of 1.5 million euros from the ERC. With the LUNAR project (Lifelong Universal lAnguage Representation), Ruiz Costa-Jussà will study a more efficient machine translation system than those offered nowadays, with similar qualities for majority and minority languages as well.

 

In particular, research will be done on improvements in the neuronal systems, which are the basis of machine translation since 2014. Despite depending on translation banks, the system based on deep learning establishes a sort of intermediate language with a mathematical nature where all the translations are operated. That streamlines the process, and by focusing on encoding and decoding the "transitory language”, it achieves a better quality. Moreover, the LUNAR project will try to find a solution to the problem of underrepresented languages ​​in text and audio databases that supply the system.

 

It should be pointed out that with LUNAR it would be possible to use the system in machine voice translation. If successful, this would be a major breakthrough in this field of research, as it is a feature that big companies have yet to develop. Ethics are a key aspect in translation systems, and this project will endeavour to report and mitigate as much as possible the biases that currently exist in machine translations..

 

Marta Ruiz Costa-Jussà has been awarded last March, and for the second year in a row, in the Google Faculty Research Awards, one of the most prestigious distinctions in computer science and engineering at global scale. Her project of the multilingual machine translation system of voice and text, ''Interlingua-based Neural Machine Translation to reduce complexity and better exploit low resources in highly multilingual environments”, was the winner.

 

Ruiz Costa-Jussà is also a researcher at the Intelligent Data Science and Artificial Intelligence (IDEAI), she is part of the Centre de Tecnologies i Aplicacions del Llenguatge i la Parla (TALP), and the Grup de Tractament de la Parla (VEU)  of the UPC. In addition to being a professor at the FIB, she also teaches at the Escola Tècnica Superior d'Enginyeria de Telecomunicació de Barcelona (ETSETB), and she is academic co-director of the Artificial Intelligence with Deep Learning postgraduate course at the UPC School. Likewise, she has worked at the Laboratoire de recherche pluridisciplinaire (LIMSI-CNRS) in Paris, at the University of São Paulo, at the Infocomm Research Institute in Singapore, at the National Polytechnic Institute of Mexico and at the University of Edinburgh.

 

The ERC, by means of the Starting Grants, aims to encourage the highest quality research in Europe through competitive funding, as well as supporting research led by experts in all fields.