There is no nice way, but you can always get the HTML (or the parse tree, depending on whether you want parsed or raw refs) and process it; references are not hard to extract. For the wikitext version, there is a python tool: https://github.com/mediawiki-utilities/python-mwrefs