In the NWO DATA2PERSON project BLISS (Behaviour-based Language-Interactive Speaking Systems), researchers from the Center for Language and Speech Technology (CLST) of Radboud University worked together with colleagues from Human Media Interaction (HMI) of the University of Twente, the company Games for Health (GfH) and the company ReadSpeaker on developing a voicebot that can communicate with people in Dutch about their daily lives and well-being. The BLISS bot can have personalized conversations about hobbies, favorite activities and interests that give users insights into their own lives, such as insights into their social network and their passions.

Our time is marked by regained interest in the region, explained as a response to immigration, globalisation, and identified with (eth)nostalgia and political populism. By contrast, scholarship on the 19th- and early 20th-century region has interpreted local colour, the region’s cultural representations and the literature which recorded the folklore and customs of its people, through the lens of nation building and nationalism. As such, studies have overlooked fundamental transnational dimensions of local colour, in its themes and representations, the circulation of local colour imagery and narratives across and beyond Europe, and its engagement with transnational audiences, through periodical reprints or as translations. Redefining the Region aims to examine these unexplored transnational dimensions of local colour, by studying media representations of regions and local colour fiction from the long 19th century in European and transatlantic frameworks. These materials provide unique case studies for transnational approaches: images and texts about European regions reached transnational audiences through emerging periodical cultures, dissemination of local colour fiction across Europe, and transcultural networks with North-American diasporic communities. Therefore, this project will yield groundbreaking knowledge about how past and present conceptions of the region are intertwined with negotiations of multiculturalism and globalisation. This will expand understandings of processes of identity/community construction. The project’s objectives are: 1) examining portrayals of regional-transnational dynamics (foreigners, emigrants) in illustrated periodicals and local colour fiction; 2) researching the transnational circulation and reception of European local colour imagery, reports, fiction; 3) exploring the reconstruction of European regions by the media and local colour writers across the Atlantic. Combining methodologies from identity, gender, diaspora/migration, reception studies, the project will additionally refine our knowledge of processes of cultural transfer and of the role of diasporic communities in cultural production and identity formation. The project will develop a searchable, digital repository, educational resources and virtual exhibition.

Deaf communication without a shared language investigated the ease with which deaf people are known to communicate across sign language boundaries. It studied both first-time dialogues of Dutch deaf signers with people from Belgium and China, as well as the performance and comprehension of sign language interpreters working for mixed international audiences. In addition, the degree of lexical overlap between sign languages was studied. Findings show that although there is some overlap between the lexicon of sign languages, this is not enough to explain the ease with which cross-linguistic communication takes place.

Europe’s recent crises have sparked Euro-scepticism and anti-immigration sentiments. Heritages of Hunger aimed to overcome divisions by reassessing education about famines (1845-1947) and developing resources to strengthen trans-European heritage consciousness. The project, led by Radboud University, Wageningen University & Research and NIOD, investigated famine legacies in textbooks, museums, and commemorations, highlighting resilience and transnational solidarity. It created educational modules, a digital exhibition, and a database of famine legacies. Surveys and interviews with educators informed the development of resources to foster historical empathy and transnational identities. The project culminated in policy recommendations and tested educational materials.

The combination of Deep Learning and Big Data has revolutionized language and speech technology in the last five years, and constitutes the state of the art in domains ranging from machine translation and question-answering to speech recognition and music analysis. These models are often now so accurate that many new useful applications are being discovered with potentially significant impacts on individuals, businesses and society. Alongside that power and popularity, new responsibilities and questions arise: how do we ensure reliability, avoid undesirable biases, and provide insights into how a system arrives at a particular outcome? How do we leverage domain expertise and user feedback to improve the models even further? In all these issues, ‘interpretability’ of the deep learning models is key. In the proposed project, pioneering researchers in the domain of interpretability of deep learning models of text, language, speech and music come together. They collaborate with companies and not-for-profit institutions working with language, speech and music technology to develop applications that help assess the usefulness of various interpretability techniques on a range of different tasks. In ‘justification’ tasks, we look at how interpretability techniques help give users meaningful feedback. Examples include fraud detection from large email collections, legal and medical document text mining, and audio search. In ‘augmentation’ tasks we look at how these techniques facilitate the use of domain knowledge and models from outside deep learning to make the models perform even better. Examples include machine translation, music recommendation, and speech recognition. In ‘interaction’ tasks we allow users to influence the functioning of their automated systems, by providing both interpretable information on how the system operates, and letting human-produced output find its way into the internal states of the learning algorithm. Examples include adapting speech recognition to non-standard accents and dialects, interactive music generation, and machine assisted translation.