I have to create a UML activity diagrams for a searching application for a semantic database. Easy to guess, users can search and select a file, but they can also browse and refine the retrieved results by clicking the categories the results belong to. They can even select a file just by browsing without actually entering any keyword. Should I create one single 'large' activity diagram showing the different possible workflows or a series of smaller ones, such as "perform a search", "browse the collection" or "select a file"?
I think, what you have mentioned is more likely to be a sequence diagram ("perform a search", "browse the collection" or "select a file"), and I would recommend you to have several sequence diagrams that you listed. My argument for this is: these diagrams are consumed by testers at design stage, so they can write tests for it.
And the major activities you can put into overall activity diagram, that will show the possibilities of what user can do at a given step/stage (if I can logged in, I can perform a search, browse the collection, etc).