The Trouble with Triples

In the Star Trek episode “The Trouble with Tribbles” the crew of the starship Enterprise encounters creatures called Tribbles. They are cute, simple creatures of mysterious origin that seem harmless at first but when they multiply the pose a big problem for the ship and the crew.

Tribbles © 1967 Paramount Pictures

Tribbles

Tribbles © 1967 Paramount Pictures

Representing and reasoning with knowledge have surprisingly similar problems. A single fact on its own is a relatively straightforward affair. A fact, such as “Tribbles are cute” can be represented with only three parts of a triple: a subject Tribbles, a predicate are and an object cute. Things can get quite a bit more difficult when there are more facts/triples: “Tribbles are round”, “Tribbles are furry”, “Tribbles originate from Iota Geminorum IV”, and other millions of facts that one could have about such a species. This is especially true when one takes into the account that the fact that knowledge can be interlinked “Iota Geminorum IV is a planet”, “Iota Geminorum IV is also known as Fafniri”, “Iota Geminorum IV is also known as Tribble Prime”.

This makes representing and reasoning with facts a non-trivial process. A system that holds all this knowledge should be able to answer a query such as “Do Tribbles originate from Fafniri?” with a yes, based on the facts “Tribbles originate from Iota Geminorum IV”, “Iota Geminorum IV is a planet” and “Iota Geminorum IV is also known as Fafniri”, even in the context of millions of other triples.

Another interesting issue with representing facts is the context of the information. To us the viewers, and initially to the crew of the Enterprise, Tribbles look like harmless and adorable creatures. To the Klingons they are an ecological menace and their mortal enemies. How such “knowledge about knowledge” is represented and used is often a challenging problem.

Various technologies have been proposed to deal with the above-mentioned issues. The Semantic Web technologies of Linked Data and Ontologies in particular have been designed around solving many of these problems. Nonetheless there is room for improvement. In the future I hope to be able to explain how these techniques can be utilized and perhaps lessen the pain points that currently surround (the use of) them.

Related