Workshop on Machine Translation for Morphologically-Rich Languages

… and Finnish in particular

• Date: 12 September
• Place: University of Helsinki, Siltavuorenpenger 3 A-Athena, 168
• Registration: click here (no fees)

The goal of this workshop is to bring together people with interest in machine translation for morphologically-rich languages and Finnish in particular. Morphological complexity and word order freedom are difficult challenges in current MT models and we would like to accelerate the development of systems that can handle languages like Finnish in a more appropriate way especially as the target language. We hope that this event will create new ideas and fosters collaborations leading to substantial improvements in MT for such languages. The following topics will be discussed during the workshop:

• Neural MT for morphologically-rich languages
• Tools and resources for Finnish (and other morphologically-rich languages)
• Evaluation of highly-inflecting languages

Workshop Program

• 09:30 – 10:00  Coffee
• 10:00 – 10:15  Welcome and workshop goals (Jörg Tiedemann)
• 10:15 – 10:30  Short introductions of participants
• 10:30 – 11:20  Current trends and tricks of the trade in NMT (Barry Haddow)
• 11:30 – 12:30  MT models for Finnish (presentations, demos, ideas)
  • MT in Helsinki/Turku/Uppsala (Sara Stymne and Jörg Tiedemann)
  • MT at Aalto (Stig-Arne Grönroos and Sami Virpioja)
  • Experiences from AbuMaTran (Antonio Toral)
  • Helsinki Neural MT (Robert Östling)
• Lunch (at Restaurant Piano, Rauhankatu 15)
• 13:30 – 15:00  Tools, data and resources
  • MT at LingSoft (Tommi Nieminen)
  • MT at Yandex (Irina Galinskaja)
  • GF and MT (Aarne Ranta)
  • Using oMorfi with Apertium and Moses (Tommi Pirinen)
  • Tools for highly-inflecting languages (Francis Tyers)
  • Mining parallel data from incomparable corpora (Filip Ginter)
• Coffee
• 15:30 – 16:30  Tuning, optimising and evaluation
  • Tuning sequence models (Turku)
  • MT quality and evaluation for Finnish (Maarit Koponen)
  • Discussion
  • • good frameworks for comparing approaches and implementations
    • tricks of the trade and practical experience in building systems
    • evaluation for highly-inflecting languages
    • collection of clean and noisy data
• 16:30 – 17:30  Ideas, plans, collaborations
  • group discussions
  • open discussions
  • hands-on + demos + crazy ideas
  • summary and plans for the future
• Drinks and food (at Kolme Kruunua, Liisankatu 5)

Local Information

The workshop is hosted at the central campus in Helsinki. The following map shows you how to get to the venue from the central railway station.

Acknowledgements

We would like to thank BAULT for funding this event!