Difficulty Prediction for Language Tests

Cognate pairs for several languages

Lisa Beinborn, Torsten Zesch and Iryna Gurevych: Cognate production using character-based machine translation, Proceedings of the 6th International Joint Conference on Natural Language Processing (IJCNLP), p. (to appear), October 2013. [PDF]

Resource download

  • Cognates for the following language pairs can be used for research purposes: en-es, en-de, en-ru, en-el, en-fa, de-cz (63.5kb)
  • The training and test data for the en-es experiments and the resulting models (19.3mb)
  • The training and test data for the other language pairs, the resulting models and some pre-processing components (88.3mb)

If you need more information, contact Dr. Lisa Beinborn