Abstract
This paper argues for the development of parallel treebanks. It summarizes the work done in this area and reports on experiments for building a Swedish-German treebank. And it
describes our approach for reusing resources from one language while annotating another language.