Instead of associating an audio file with a web document, why not associating it with web data? I always thought an audio player should act as an aggregator, and allow me to display the aggregated information in any way I like. For example, I may aggregate information about chords and want to display that. I may aggregate geographical information and plot my collection on a map. etc.
I did a small implementation of that some time ago [1]. Basically, I put a RDF URI in the WOAF field of tracks within my collection. Then, the collection management tool aggregates information by crawling from these URIs. Easy. Then, I can play with all this data in whatever way I like – I have a tailored database for my music collection. “Create me a playlist of musical works composed in the 30s and performed in NYC in the 70s”, “Create me a playlist of artists married to each others involved in my collection”, “Give me two tracks with a similar chord progression”, “Sort me hip-hop artists in my collection by murder rates in their city” :-) All that are the sort of interaction we can do right now by considering a media player as an aggregator.