Digital libraries are a crucial infrastructure for researchers to find and access resources. For example, researchers in the digital humanities and computational literary studies make frequent use of digital libraries to access relevant materials, such as text corpora, which they then annotate. These materials are often protected by copyright and are typically released under licensing terms that either prohibit redistribution entirely or impose significant limitations on how they can be shared. We present and evaluate an approach to separate annotations from the underlying text and create a fingerprint, which cannot be used on its own to recreate the original text and can therefore be shared. This fingerprint can be used to merge the separated annotations with another version of the original text. Our framework can easily be adapted to support different file formats and can be integrated into digital libraries to simplify sharing of annotations of copyrighted texts. The code is publicly available under the Apache License 2.0 at https://hu.berlin/sisc .