#1: Pioneers

We are pleased to announce the first MT Meetup in Zurich. The event will take place on 19th April 2018. 50 seats available, please RSVP on our meetup site. Two pioneers in the field of machine translation will be speaking: MIRKO PLITT, Head of Technology at Translators without Borders, and RICO SENNRICH, Assistant Professor in Machine Learning at University of Edinburgh.

Schedule:

19:00 Welcome, Intro
19:15 Talk 1: Mirko Plitt - "Post-Editing: On Productivity and Quality"
19:50 Talk 2: Rico Sennrich - "Neural Machine Translation: Breaking through the Performance Ceiling"
20:30 Apéro and Networking

------------------------------------------------------
Mirko's Talk: "Post-Editing: On Productivity and Quality"

I will review how the application of machine translation has evolved along with the underlying technology and technology in general, within the constraints of the translation industry's equally evolving setup. In my talk, I focus particularly on the adoption of post-editing methods as part of large-scale translation processes in recent years. I will show that the introduction of the industrial concept of productivity was instrumental to the successful adoption of MT in these contexts as it allowed to shift the debate away from the highly problematic notion of translation quality.

About MIRKO:
Mirko was among the very first to leverage machine translation for industry-level content translation. In a seminal user study on post-editing productivity (Plitt and Masselot, 2010), he showed that machine translation can save translators up to 40% of time in several language pairs. His work addressed many of the day-to-day problems that translators face when working with potentially erroneous system output, most notably including domain-specific terminology. Before founding Modulo Language Automation in 2013, a startup which developed a novel approach to the automatic assessment of post-edited translations, Mirko was a Senior Language Technology Manager at NASDAQ-listed US software company Autodesk. Since 2016, he also serves as Head of Technology at Translators without Borders, a non-profit organization offering language and translation support for humanitarian and development agencies, where his work has expanded to under-resourced languages and spoken translations for illiterate and visually impaired audiences.

Rico's Talk: "Neural Machine Translation: Breaking through the Performance Ceiling"

Machine translation is a challenging NLP task that involves both natural language understanding and generation. In my talk, I will present our recent research addressing limitations in neural machine translation. We show that we can encode vocabularies of infinite size through a closed vocabulary of subword units, while maintaining computational efficiency. We also explore semi-supervised learning for neural machine translation, and show substantial gains in translation quality with a simple, widely applicable algorithm for including unsupervised training data.

About RICO:
When most researchers and practitioners still deemed neural machine translation a hype, Rico built Nematus. It soon became one of the most successful toolkits for language translation with neural networks to date. In the prestigious news translation task at WMT 2016 and 2017, his translation systems outperformed competitors – including Google Translate – by a large margin. He used a clever trick to solve the problem of unknown words in neural machine translation, splitting up rare words into subword units. Among many other methods introduced by Rico, the use of subword units is now standard practise in building state-of-the-art machine translation systems in research and industry settings alike. Rico obtained a PhD in Computational Linguistics from the University of Zurich in 2013, and has been appointed Assistant Professor in Machine Learning at the University of Edinburgh in 2017.