10:00 - 10:10
Welcome (Luc Meertens, Consortium)
10:10 - 10:30
How does this workshop fit into DG CNECT's roadmap? (Philippe Gelin, DG Connect, European Commission)
10:30 - 11:00
Reacting to Information Needs in a Global Pandemic: How to Build Multilingual Information Access Systems for Covid-19 (Nicola Ferro, University of Padua)
11:00 - 11:30
Multilingual Mobile Health Along the Humanitarian Aid Chain: a Tase for Clinical Malaria Diagnosis in Ghana (Celia Rico Perez, Universidad Europea de Madrid)
11:30 - 11:45
11:45 - 12:15
Using Language Technology to Enable Two-Way Communication in Humanitarian Assistance (Eric Paquin, Translators without Borders)Presentation
12:15 - 12:45
Rapid Development of Competitive Translation Engines for Access to Multilingual COVID-19 information (Alberto Poncelas, Dublin City University)
12:45 - 14:00
14:00 - 14:30
Towards a Cross-Media Approach for Communication During Crises (Gerhard Backfried, Sail Labs)
14:30 - 15:00
NLP for Teamwork Support in Robot-Assisted Disaster Response (Ivana Kruijff-Korbayova, DFKI)
15:00 - 15:30
Cross-lingual Generalization, Alignment and Applications (Junjie Hu, Carnegie Mellon)
15:30 - 16:00
Outro (Luc Meertens, Consortium)
The overall objective is to contribute to the development of CEF Automated Translation as a “multilingualism enabler” for CEF DSIs, online services linked to CEF DSIs and other relevant public online services. The specific objectives are (i) to gather information on additional needs of the CEF DSIs and public services; (ii) to analyze the range of services which could extend CEF Automated Translation and (iii) to support CEF DSIs and related systems with a view to maximizing their use of CEF Automated Translation services.
The current Covid-19 outbreak has been hitting our society in many ways and one of the critical challenges is how to provide access to the growing amounts of information in an efficient, effective, and trustworthy way. Indeed, in the current Covid-19 crisis, as in many other emergency situations, the general public, as well as many other stakeholders, need to aggregate and summarize different sources of information into a single coherent synopsis or narrative, complementing different pieces of information, resolving possible inconsistencies, and preventing mis-information. This should happen across multiple languages, sources, and levels of linguistic knowledge that varies depending on social, cultural or educational factors.
Unfortunately, our information access systems are not tailored and optimized (yet) on the specific Covid-19 domain in many respects, for example, for accessing general public source, or scientific literature, or regulatory and policy documents, just to name a few. And this is even more challenging when multiple languages and low-resourced languages come into play.
We will present and discuss what the Covid-10 MLIA @ Eval initiative (http://eval.covid19-mlia.eu/) is doing to face these issues. Covid-19 MLIA @ Eval is a community effort to boost the development of (language) resources and Multilingual Information Access (MLIA) systems specifically tailored on Covid-19. In particular, we organise evaluation tasks that steer the development of systems and resources in the following areas: information extraction, multilingual semantic search, and machine translation.
Mobile Health (mHealth) is a rapidly developing field. It is estimated that over 100,000 mHealth apps are currently available on the market. mHealth contributes to the empowerment of patients as it allows for them to manage their health more actively. It also supports healthcare professionals in treating patients more efficiently. Along the Humanitarian Aid Chain, the multilingual needs of stakeholders —donors, NGOs, local partners and beneficiaries— are catered for unevenly, depending on the available resources and the crisis urgency. The use of mobile apps in this context provides a useful means of implementing language tools for crisis response. Such is the case of clinical malaria diagnosis in Ghana. By creating a multilingual mobile app for the diagnosis of febrile processes such as malaria, we can address a double challenge: 1) responding to the problem of communication with the patient in less resourced languages; and 2) contributing to stop the spread of infectious epidemics in endemic areas by collecting key information that is not otherwise easily accessible.
Many, if not most, of the 204 million people in need of humanitarian assistance in the world today cannot communicate with aid workers. They can't be understood and can't understand the life-saving information that they need because it's not in their language. We know technology can help bridge the language gap andthis is why Translators without Borders has invested in research, development and deployment of language technology solutions. This presentation will cover current use of technology and which ones we are excited about for the future.
Alberto Poncelas, Dublin City University (Rapid Development of Competitive Translation Engines for Access to Multilingual COVID-19 Information)
Every day, more people are becoming infected and dying from exposure to COVID-19. Some countries in Europe like Spain, France, the UK and Italy have suffered particularly badly from the virus. Others such as Germany appear to have coped extremely well. Both health professionals and the general public are keen to receive up-to-date information on the effects of the virus, as well as treatments that have proven to be effective. In cases where language is a barrier to access of pertinent information, machine translation (MT) may help people assimilate information published in different languages. Our MT systems trained on COVID-19 data are freely available for anyone to use to help translate information (such as promoting good practice for symptom identification, prevention, and treatment) published in German, French, Italian, Spanish into English, as well as the reverse direction.
Gerhard Backfried, Sail Labs (Towards a cross-media approach for communication during crises)
An increase in the number and intensity of disasters paired with a growth in the diversity and variety of traditional and social media reporting about them provide a strong case for an approach accommodating the wide variety of platforms and media in a cross-media approach. Such a cross-media approach should account for the following principles:
Media and platforms are frequently used in combination, linking actors and contents across several platforms. The complete picture can only be obtained by the combined view rather than by individual views from solitary platforms. These are by necessity only a small window into a larger communication landscape. Based on experiences made during a research project on floods in Europe and an operational system for the monitoring of open sources we argue that traditional and social media complement each other and should be used in combination rather than in isolation for use during disasters and crises.
Air and ground robot systems are increasingly used in disaster response operations, e.g. for site reconnaissance. We have been working with first responders for almost a decade to investigate teamwork and teamwork support in robot-assisted disaster response in realistic scenarios. Our specific focus is on developing language technology for interpreting the verbal communication among the human team members with the goal to extract mission knowledge as the mission unfolds, i.e., the goals, the tasks within the human-robot team and the state of their execution. We also develop mission assistance tools based on mission process models. This includes mission management-support and mission documentation. I will introduce the components of our system and illustrate some of the challenges for NLP with practical examples.
While text on the web is an invaluable information source, this text is not available in large quantities for most languages in the world. It is even difficult to ask native speakers to annotate text in most languages for training individual machine learning models. With recent advances in multilingual machine learning models, we are able to transfer knowledge across languages in one single model, and apply the model to deal with text written in more than 100 languages. However, a benchmark that enables the comprehensive evaluation of such models on a diverse range of languages and tasks is still missing. In this talk, I will focus on analyzing cross-lingual generalization effects in these models, and propose methods to improve the performance in real applications. Specifically, I will start with introducing Cross-lingual TRansfer Evaluation of Multilingual Encoders (XTREME) benchmark, a multi-task benchmark for evaluating the cross-lingual generalization capabilities of multilingual models across 40 languages and 9 tasks. Secondly, I will show that a compact multilingual model trained on parallel translation text can align multilingual representations, performing on a par with or even better than much larger models on NLP tasks such as sentence classification, and retrieval. Finally, I will present our recent translation initiative for COVID-19, a multilingual translation benchmark in 35 different languages, in order to foster the development of tools and resources for improving access to information about COVID19 in these languages.