The wiki forms a graph. This same MiniWiki engine may serve as the backbone of a self-organizing map, a neural network, a collaborative filter, a matrix transform, a game of life.
The links in content hypertext form are used to connect short writings to one another. The same links could be provided with numeric or encoded values to be operated upon by a feedback machine learning algorithm. Perhaps a blend of content and link traversing iteration would be more beneficial to knowledge extraction and creation.
For example, if each content page were a stock chart, a bot could crawl, updating links between those stocks with price correlations.
Human feedback may be more valuable, at least to begin with. A web could be created with a number of images, and using other links for images, relationships between pictures could emerge, to show the images that fit or fail a criteria.
Entering choices into a decision tree is another use--gathering information along a "twenty questions" route will deliver at the end a narrow group of possibilities, or even a single best fit.