Straight-Up Triples (Triple Extraction - Part 1)
CraigTrim 110000G799 Visits (4559)
A transitive verb demonstrates an action that passes from one entity to another. We might say that the first entity (the subject) commits the action on the second entity (the object). The action is what links the subject to the object.
This connection between a subject and an object is considered a triple. The action passes over from the subject to the object. Transitive verbs lend themselves well to triple extraction from unstructured text.
The performer of the action is not always the subject of the verb. If the action performer occurs after the transitive verb, this is considered a passive sentence construction. It is considered good form to place the action performer before the verb. Yet there is no guarantee a sentence will be formed this way.
Both sentences express the same meaning. In the first sentence, coyotes actively initiated the action. In the second sentence, the passive sheep are emphasized. They are done by the action, and this makes sheep passive.
The distinction between active and passive sentence constructions in triple extraction is not a strict necessity. Given a parser capable of recognizing the verb patterns, “Hamlet was written by Shakespeare” and “Shakespeare wrote Hamlet” both render conceptually identical triples:
Although the use of predicate properties is beyond the scope of this article, use of the owl:inverseOf property would enable the system to consistently interpret the extracted information:
Or given the appropriate domain and range assigned to a predicate, and with appropriate annotative capability such that Hamlet is known to be a type of “Book” and Wm. Shakespeare a type of “Author”, we can simply find the predicate with matching domain/range values, and plug this information in. Such a predicate would likely be named “authorOf”, and both sentence fragments above then render: