Universal Dependencies and Morphology for Hungarian - and on the Price of Universality

Abstract

In this paper, we present how the principles of universal dependencies and morphology have been adapted to Hungarian. We report the most challenging grammatical phenomena and our solutions to those. On the basis of the adapted guidelines, we have converted and manually corrected 1,800 sentences from the Szeged Treebank to universal dependency format. We also introduce experiments on this manually annotated corpus for evaluating automatic conversion and the added value of language-specific, i.e. non-universal, annotations. Our results reveal that converting to universal dependencies is not necessarily trivial, moreover, using languagespecific morphological features may have an impact on overall performance.

Efficiency Analysis of Deeplearning4J Neural Network Classifiers in Development of Transition Based Dependency Parsers

“…Several different notation systems have been developed to provide the mentioned annotation, including UD (Universal Dependencies) [14] which aim is to provide a general-purpose language independent solution. Among others there is a Hungarian corpus available within the UD project [15]. Also, there is a much larger Hungarian text corpus that is following the file format that was defined in the UD project, however the annotation label, the morphological description and the POS tag set are different.…”

One format to rule them all – The emtsv pipeline for Hungarian

“…Formerly in e-magyar, the model behind emDep was trained on POS tags and morphosyntactic features converted by DepTool. Consequently, the model had to be replaced with one trained on Szeged Treebank with UD tags (Vincze et al, 2017).…”

Creation of a corpus with semantic role labels for Hungarian

“…As a starting point, we chose the Hungarian subcorpus (Vincze et al, 2017) of the Universal Dependencies (UD) corpus (Nivre et al, 2016) consisting of 1800 sentences (42000 tokens) of mainly newswire text in order to put the annotation schema we propose in a context that can be interpreted at an international level. The UD corpus contains texts in many languages annotated with morphosyntactic and dependency-based syntactic analysis using unified principles and categories.…”