Licenses of digital resources describe rights and duties for users. If the licenses are expressed in natural language, as it is frequently the case, it is hard to reason and verify the license compatibility to specific uses. We propose an automatic end-to-end workflow for extracting Open Digital Rights Language (ODRL) representations from textual license documents. This process uses AMR semantic representations as an intermediate; it adapts a tool that performs a semantic transduction analysis, using formal rules. This work focuses on deontic modalities expressing the permissions and obligations of the user. We provide a proof of concept and discuss experiments.