Background

Mainstream language technology is dominated by neural networks. Neural networks provide powerful tools for language modeling and there are now an increasing amount of university classes, seminars and workshops about it.

Good to read

• NN for NLP course at Carnegie Mellon University
• NLPL winter school
• also interesting: Recent trends in deep learning

• Grave et al 2018: Learning Word Vectors for 157 Languages
• Modelling Natural Language, Programs, and their Intersection
  • slides

• MT baseline, cf. our rule-based MT
• Stavekontrollforslag, cf. our work on proofing
• Missinglist-komponent

Vi har språklege data (i Korp nå, men vi har mer):

• sme: over 32M tokens, nesten 3M setninger
• nob-sme: nesten 2.5M tokens, mer enn 150K setninger