To remove confusion, one should be clear about the context of the categories, not reduce the list of options. The first data set "Datasets, Posters, Presentations, Publications, Software" is already confusing because it looks like "everything". If one selects that, you're just presented with a list of names of categories, no description of the categories and what is appropriate for them. Assuming that the user will understand from the name of the category what should go into the category is a bit of a stretch.
Of course the metadata needs to be as complete as possible, but you also want to guide the visitor. If you're worried about exploding the list of categories, why not remove the distinction between a "Sci-GaIA" presentation and an "other" presentation - this information is itself in the metadata, if it's properly compiled. On the other hand, a "presentation" is not a "lecture" - I think they are sufficiently different to warrant an entirely different category.
Also, what would a visitor search for ? Keyword search is likely to yield very different results based on who is searching, and a search with terms like "big data" for example will yield hits from across the categories (datasets, lectures, presentations, analyses, etc). Sure, the user can then read the abstract and tell a "big data" lecture from a course apart from a "big data" presentation at a conference, but it would be far quicker to see that one is published under "OER" or "Educational Material" and the other is published under "Conference proceedings" or similar, no ?
So, I still think some changes to the categories are warranted.