RDF Serialization is abstruse, and there have been a lot of criticisms of it in the XML community. Recently, Tim Bray tried to express RDDL in RDF and decided it would be easier to redo RDF (look on the XML-DEV and W3 Tag email lists).
While I feel RDF is harder to understand than it needs to be, I think its good to take a step and think about XML's contribution to difficulty of human comprehension of RDF. XML is the very poorly chosen for human comprehension and editing, although its easier than using binary files. One only has to do a little XSLT coding. The effort and number-of-lines of code saved by a using such a high-level language is frequently lost in dealing with lines and lines of decoration required by using XML Tagged Text serializatoin of the infoset.
Python and Haskell both have much cleaner syntaxes than languages based on start and end tokens (like xml, which has openening and closing tags, or C, which has opening and closing braces). In Python and Haskell, people are not spending nearly as much time on syntax decoration (lines of code required just to make the computer happy) and they are visually much easier to read than C-like languages or XML because they are based on indentation.
A serialization of XML called SOX shows how much easier a non-tagged text markup language can be to read compared to tagged-text: http://www.langdale.com.au/SOX/
The difficulty in human comprehension of XML-serialized RDF very arise more from XML structure than the RDF representation within the XML infoset. XSLT programs provide other examples of how very simple programs are hard for people to read.