Automatic Conversion of the Basque Dependency Treebank to Universal Dependencies

This work describes the process of automatically converting the Basque Dependency Treebank to Universal Dependencies (UD). Our objective is to develop a set of conversion rules that will automatically transform the original treebank to UD. Basque is a morphologically rich and agglutinative language, which presents different challenges for the conversion from the initial annotation scheme to UD. We will illustrate the steps pursued and the main difficulties we have encountered. As a main conclusion we can say that, although the Basque original treebank was in accord with many UD guidelines, the process was not trivial, converting around 80% of the tokens.