Building a UD treebank using existing resources from related languages: the case of Galician

This paper presents a novel strategy for creating a Universal Dependencies (UD) treebank of a low-resource language. The method consists of adapting and combining different UD treebanks from related varieties in order to train a parser for the target language. More precisely, the paper explores the influence of three different levels for the selection and adaptation of the source treebanks: (i) the relatedness of the linguistic varieties, (ii) the adaptation of features based on lexical and spelling data, and (iii) the agreement in annotation criteria between different treebanks. The proposed strategy allowed us to train a parser for analyzing, with promising results, a small Galician corpus without previous availability of labeled data for this language. After a few bootstrapping iterations, we obtained a UD gold-standard corpus, used for proving the effectiveness of the proposed method.