Automatic Morphological Enrichment of a Morphologically Underspecified Treebank

Sarah Alkuhlani, Nizar Habash and Ryan Roth

In this paper, we study the problem of automatic enrichment of a morphologically underspecified treebank for Arabic, a morphologically rich language. We show that we can map from a tagset of size six to one with 485 tags at an accuracy rate of 94%-95%. We can also identify the unspecified lemmas in the treebank with an accuracy over 97%. Furthermore, we demonstrate that using our automatic annotations improves the performance of a state-of-the-art Arabic morphological tagger. Our approach combines a variety of techniques from corpus-based statistical models to linguistic rules that target specific phenomena. These results suggest that the cost of treebanking can be reduced by designing underspecified treebanks that can be subsequently enriched automatically.