Eurovoc has been mentioned as a candidate as a controlled multilingual vocabulary to classify datasets. It may be a good idea to research alternatives before deciding on Eurovoc. Some of the drawbacks with Eurovoc are:
1. Complexity: the person who has to enter the topic for a dataset has a lot of topics to choose from. They may opt for high level topics which will reduce precision in the aggregated catalog. Also, a software solution for a smaller catalog needs to import Eurovoc and provide som sort of browsing interface.
2. Mapping between topics: the SKOS version of Eurovoc seem to be designed with one URI per conept per language instead of concepts with multiple labels. This requires multiple operations to cross reference topics.
Any reference implementations should study how these issues affect usability.