Workshop on Crisis Machine Translation (“Crisis MT”) at MT Summit 2019

We are pleased to make a call for submissions to the First Crisis MT Workshop, which will be held on Tuesday, 20th August, 2019 at the MT Summit 2019 Conference in Dublin City University.

Workshop Context

Timely and accurate information is known to save lives in crisis or disaster settings (Fischer 2008), and is now considered to be a human right (Greenwood et al. 2017; O’Brien et al. 2018). Crises (broadly understood) are known to have cascading effects (Pescaroli and Alexander 2015), to be transboundary, cross-cultural and cross-linguistic. However, the role of translation as an enabler for crisis communication, in all stages of a crisis (preparedness, resilience building, response, recovery) is rarely considered, and the role of technologies such as machine translation, to aid multilingual crisis communication has received even less attention. Some exceptions to this were the work of Lewis (2010) and Lewis et al (2011) in response to the Haiti earthquake and the work of Translators without Borders (Zetsche 2017). More recently, researchers in the EU-funded INTERACT project have been tackling various aspects of this issue by testing pivoting techniques for MT (Liu et al 2018; Silva et al 2018), creating “How to” course content for citizen translators and post-editors, and considering and commenting on the ethical dimension of translation and translation technology in crisis response. MT for crisis settings brings with it a considerable number of challenges, such as data sparsity, translating between “unusual”, or low-resource, language pairs, for languages that have sometimes never been heard of in the commercial world of MT, with very little time, potentially no power or internet connections, and with potentially negative consequences if the output is inaccurate. The proposed workshop seeks to address the many challenges of crisis MT and is open to all researchers interested in this topic, from areas such as NLP, translation studies, disaster and humanitarian response.

Call for Extended Abstracts

We call for extended abstracts (400–500 words). Abstracts must follow the structure below:

1. Title of talk; name and affiliation of contributors

2. Topic covered (from list below or additional topics)

3. Problem statement (what is the issue being tackled in the proposed talk?)

4. Design/method/approach (how did you investigate the topic?)

5. Findings (what findings do you propose to present?)

6. Contribution (what is the contribution to the field of crisis MT? Note that there has to be clear and direct applicability to the topic)

Topics can include, but are not limited to:

Machine translation for low-resource languages that may be needed in crisis, emergency or disaster settings, including topics such as:Data curation for “readiness” or “preparedness” for disasters; Challenges of rapid response MT engine creation and how these might be tackled;Data sparsity challenges and how they can be overcome, e.g. domain tuning, pivoting, hybrid approaches, artificial data, back translated data etc.;Evaluation challenges;MT for mobile devices, including feature phones;MT for offline use;Multimodal MT and how this might be of benefit to crisis settings; Quality estimation or other techniques for triaging MT output in high-risk settings; Quality assessment and quality assurance for MT in crisis settings;Ethical challenges of using MT and related technologies for timely and accurate crisis communication and how these might be overcome or tempered;Genres of texts and communication that are included in crisis settings and the various challenges they introduce for MT (e.g. non-standardised writing systems, “textspeak”, typos, use of emoticons in text, cross-domain communication including medical, technical and general communication etc.);MT for social media content in crisis settings;Text to speech and vice versa with MT in the loop for crisis response;Terminologies and their management and use in crisis translation settings;Citizen post-editing;Training for MT usage in crisis settings;Related translation technologies and their usefulness for crisis response (e.g. translation memory, terminology tools);Simplification of crisis communication content and whether this has any impact on MT output;Authorship, copyright and licensing of data needed to build crisis MT systems;Best use of web-crawled monolingual, and parallel/comparable data for crisis MT systems.

Important Dates

Friday, 03rd May 2019: Submission deadline for extended abstracts (400-500 words)Friday, 31st May 2019: NotificationTuesday, 20th August 2019: Workshop

Workshop Format

This will be a full-day workshop composed of a keynote address (tbc), accepted oral presentations, and an interactive, participative “think-in” on how the challenges raised might be tackled in the future. All participants should be willing to fully engage in the interactive session.

Submission Instructions

Submit your 400-500 word abstract in PDF format to the following email address by 12 midnight GMT on 03 May: crisismtworkshop2019@gmail.com

Please ensure that the abstract is structured using the headings listed above.