| Article ID | Journal | Published Year | Pages | File Type |
|---|---|---|---|---|
| 4961868 | Procedia Computer Science | 2016 | 8 Pages |
Romanian is known for having a rich inflectional morphology due to the multitude of inflectional ending categories as well as due to stem alternations (apophony). This richness poses difficulties both for second language learners attempting to acquire the many variations, as well as for researchers attempting to describe and incorporate them in Natural Language Processing pipelines. While many studies in Computational Morphology have proposed methods to automatically learn word inflection or segmentation, very few have focused on learning apophony, specifically vowel and consonant mutation in the stem, a phenomenon typical for Romanian inflection. In this paper, we investigate the applicability of recent strategies proposed for the Romanian verbal domain 1,2 to the Romanian nominal domain.
