> It's especially interesting with good case studies for how to model an advanced tag system in a graph database
I wouldn't accuse it of being a good tag system, nor a true graph database, but one thing to look at is Semantic MediaWiki. It's a MediaWiki extension which takes Categories as a starting point, and extends it quite far with e.g. relations and key-value pairs.
One interesting feature of Semantic MediaWiki is called "Concepts" which are essentially "computed tags." They can be used in place of Categories in most places, but while Categories are set by editors on a piece of content, Concepts are defined by a query against Categories or other properties. This can help bridge gaps between different types of tags that represent different ways of thinking about the content.
I wouldn't accuse it of being a good tag system, nor a true graph database, but one thing to look at is Semantic MediaWiki. It's a MediaWiki extension which takes Categories as a starting point, and extends it quite far with e.g. relations and key-value pairs.
One interesting feature of Semantic MediaWiki is called "Concepts" which are essentially "computed tags." They can be used in place of Categories in most places, but while Categories are set by editors on a piece of content, Concepts are defined by a query against Categories or other properties. This can help bridge gaps between different types of tags that represent different ways of thinking about the content.