From Word Alignment to Machine Translation via Superlinks
Lars Ahrenberg, Håkan Jönsson · DSpace repository (University of Tartu) · 2001
This paper presents a data-driven lexicalist framework for machine translation based on alignment.In addition to word alignment that provides the word correspondences of source and target language, the system uses classifications of the correspondences based on superlags.We lirsL give an overview of the framework, pointing out its fundamental concepts and then report some results from a pilot implementation and evaluation on the ATTS domain.