In a loose analogy to the distinction between hard and soft links that is sometimes made when referring to file links, one may perhaps distinguish between loose and tight links in a knowledge base. The contrast may perhaps also be characterized as one between "indirect" links and "direct" links. Any application that relies on keywords, tags or categories to connect different entries in its database, may be said to rely on indirect or loose links. An application that relies mainly on actual references of one item to another, like a Wiki (or other hyper-textual applications) may be said to rely on direct or tight links.
I have argued before that just relying on indirect links is ineffective (and have criticized a previous incarnation of Luedecke's Zettelkasten for doing this, since it claims to have been conceived after Luhmann's system which relied almost exclusively on direct or hard links).
There is obviously no reason why a wiki application that relies mainly on direct links cannot also utilize the loose links that tags or categories provide. In fact, the two methods are not contradictory but complementary. They offer different views of the same data, possibly opening up new perspectives.
Both types of links or connections depend largely on deliberate input by the maintainer or user of the knowledge base. She has to make the link or assign the category (or keyword) explicitly.
There is an even "looser" connection opened up by an algorithm that computes which topics are "like" the topic under consideration. This kind of connection does not need special deliberation by the user. It adds another view or perspective (and one that will perhaps increase "serendipity").
1. See ConnectedText 5.