A new MLLP research project, “Multisub: Multilingual subtitling of classrooms and plenary sessions” (2019–2021), has been selected by the Government of Spain for funding under the 2018 Research Challenges call of the Societal Challenges R&D&I Programme 2017–2020.
The Machine Learning and Language Processing research group (MLLP) of Universitat Politècnica de València (UPV) has been for years leading research and innovation projects at the EU and Spanish levels. The Spanish Ministry of Science has now selected the MLLP’s 3-year “Multisub: Multilingual subtitling of classrooms and plenary sessions” (2019–2021) for funding under the 2018 Research Challenges call of the Societal Challenges R&D&I Programme 2017–2020.
Open Education (OE) is being fostered by governments and international organizations as the best way of modernizing education. The relevance of Open Educational Resources (OER) in OE is stated in the Paris OER Declaration 2012 and in the Ljubljana OER Action Plan 2017. The development of multilingual OER appears as an important social challenge to promote a wider access to education at all levels, contributing to social inclusion and allowing inter-cultural knowledge sharing. Similarly, Parliamentary Openness (PO) is also being promoted by parliamentary monitoring organizations (PMO) in order to increase the openness of parliamentary work and enhancing citizen participation. The Declaration on Parliamentary Openness (2012) also highlights the relevance of multilingualism in PO for countries with more than one official language.
Universities worldwide are beginning to see benefits in producing OER as lecture recordings to support teaching and learning activities, which are typically published in online learning platforms and are becoming even more popular among students. In the same way, parliaments are using even more video as a powerful way to bring the attention of society towards the work of parliaments, and are increasingly making live streams and recordings available allowing to search for parts of plenary sessions or speeches.
The main aim of this project is to further improve the state of the art in Automatic Speech Recognition (ASR) and Statistical Machine Translation (SMT) to deal with these kinds of audiovisual collections, addressing: i) channel variability, noise and reverberation; ii) far-field speech recognition; iii) speaker diarization; iv) multispeaker recognition; v) on-line speech recognition; vi) neural machine translation; vii) translation of sentences containing ASR errors.
The technology developed will be integrated into Opencast and Transparency Portal platforms to enable real-life impact.
The MLLP is excited to start working on Multisub along with their other current projects, which include the EU project X5gon and technology transfer contracts with several universities and companies in the EU and the US, in addition to the provision of automatic multilingual subtitling services to our own Universitat Politècnica de València. Be sure to visit our Projects section to read more about Multisub and other MLLP projects, current and past.