Or, of course, just XPath 1.0 and a bit of shell work:
$ xmlstarlet sel -t -m "//@audience | //@platform | //@props" -v "." -n
ditaDir/*.dita ditaDir/*.xml | sort | uniq -c | sort -nrs

________________________________
Content is a bunch of dita topics of varying lengths and complexity -
technical documentation.  I am trying to get a distinct list of all the values
in those attributes across all the content and possible elements.