Multi-task, Multilingual, Multi-modal Language Generation
Foster knowledge exchange by sharing of resources including semantic annotation guidelines, benchmarking corpora, machine learning and alignment tools.
Create multimodal and multilingual benchmarks for NLG involves experimenting with automatic mapping between existing resources, crawling of Web data, definition of annotation guidelines and launching of crowdsourcing campaigns for bigger datasets (also as games-with-a-purpose).
Facilitate interactions, collaborations, knowledge building and dissemination between Action participants via online tools, as website, blogs, downloadable publications.
Promote the generation of novel ideas and introduce the new joint Multi3Generation discipline to other researchers.
Provide opportunities for joint research projects by Action members on multi-task, multilingual and multimodal processing during exchange visits of Early Career Investigator, and other activities that encourage young researchers to establish links with industry and more senior academics.
Disseminate the results of the Action through conferences, scientific and industrial gatherings, which will have substantial impact in the participating countries and beyond.
Create synergies between participants via joint publications in books, journals and conferences; reports from working group meetings and training materials from training schools.