I've been exploring semantic technologies from the cold and incredible practical angle of "my own needs" :-). The RDF model offers a lot even when keeping data for myself: I can use RDF to infer new things from existing data and visualize relationships. An RDF triplestore can be used to store pretty much anything, and works well for CRUD. RDF is about graphs, so a triple store can be used to model documents too, since trees are a subset of all graphs.
RDF4J has really nice documentation and supports a lot of modern features. The getting started guide is great [1], and there's also a super friendly gitter room to ask questions [2].
1: https://rdf4j.org/documentation/tutorials/getting-started/