On 31 March 2022, almost 300 participants from all over the world came together for the 6th ELRC Conference, held under the motto “Think big – For Europe’s multilingual future”. The virtual event was opened by Philippe Gelin, Head of Sector Multilingualism at the European Commission, who explained that even though multilinguality is close to the heart of the European Commission and one of its key values, it can also create language barriers which hinder the free exchange of ideas and information. Also, he pointed out that the majority of Europeans does not speak a second language well enough to e.g., sustain a conversation, watch a movie or sign legal documents – and this is where Artificial Intelligence (AI) and Language Technology (LT) can be used to support intercultural communication in more and more situations.
Against this background, the 6th ELRC Conference focused on the latest European solutions and activities related to
Large Language Models, multimodal language data, and low-resource neural machine translation (NMT) and put a spotlight on the European Language Data Space, which is foreseen as part of the freshly introduced Digital Europe Programme. Taking the present situation in Ukraine into account, the conference also supported the “Crisis response without borders” initiative and included a discourse on using LT for fighting disinformation.
After lively discussions and exchanges, Andrea Lösch, the ELRC project manager and conference moderator summarised the key outcomes of the day:
• Large language models will outperform and gradually replace other AI solutions
Large language models are critically important to ensure Europe’s Digital Sovereignty, which is why they need to be a substantial component of future data collection efforts.
• Multimodal language data offers great opportunities
Making multimodal language data available offers great opportunities for the development of language-centric AI. Therefore, future data collection efforts need to account for language data of all types – and not be limited to textual data.
• LT plays a key role in detecting fake news
Language technologies can play a central role in helping detect fake news, thus slowing the intense flow of disinformation.
• CEF & Horizon 2020 projects are already successfully bridging language barriers
Numerous CEF and Horizon 2020 projects are investigating and/or already using a variety of techniques for bridging language gaps. This ranges from using cross-lingual embeddings to support less-represented languages in European News Media or promising attempts to bridge the gap between academic NLP research and industry, up to the successful integration of eTranslation to an AI-based chatbot technology.
Those conclusions could only be drawn thanks to the interactive exchanges between the speakers and the audience, so a big thank you goes to everyone who attended the 6th ELRC Conference and contributed to this successful event. With an overall rating of 4.6 out of 5 stars, the success of the conference was also reflected in the participants’ direct feedback.
You missed the 6th ELRC Conference? No worries!
You can simply catch up by browsing through the presentations on the ELRC website or by watching the conference recording on Youtube. A summary of the key contents and outcomes can also be found in the conference report. And of course, further ELRC events will follow – check them out here.