SZTAKI HLT | Unsupervised morphology learning

Unsupervised morphology learning

The eventual goal of this project is the unsupervised learning of morphology.

Hungarian exhibits rich agglutinative morphology with vowel harmony and a large number of noun cases.

Our current experiments include supervised morphological segmentation using recurrent and convolutional neural networks and morphological reinflection using sequence-to-sequence models.

Project leader