A Syntax-first Approach to High-quality Morphological Analysis and Lemma Disambiguation for the TüBa-D/Z Treebank

Morphological analyses and lemma information are an important auxiliary resource for any treebank, especially for morphologically rich languages since such information is a useful precondition for any task that needs to link surface forms to semantic interpretation (either through wordnets or distributional measures). In contrast to common practice in parsing, the method used in the TuBaD/Z treebank uses syntactic information for the morphological and lemma disambiguation. We argue that this approach has an advantage in the context of treebanking since many ambiguities in morphology and lemmas can be eliminated given the syntactic context.