This article considers the task of automatically inducing role-semanticannotations in the FrameNet paradigm for new languages. We propose a generalframework that is based on annotation projection, phrased as a graphoptimization problem. It is relatively inexpensive and has the potential toreduce the human effort involved in creating role-semantic resources. Withinthis framework, we present projection models that exploit lexical and syntacticinformation. We provide an experimental evaluation on an English-Germanparallel corpus which demonstrates the feasibility of inducing high-precisionGerman semantic role annotation both for manually and automatically annotatedEnglish data.
展开▼