Open information extraction is a new paradigm, which emerged to cope with the scenario where the number of relations to retrieve is too large or their type is unknown. The topic is of particular interest for security applications such as the monitoring of terrorist networks, relying heavily on one's ability to discover undisclosed connections between individuals, organizations, events or locations. This paper describes an approach developed to identify relevant relationships from textual data. Relations are understood as associations between people, organizations, locations and events, and are extracted by using a text mining algorithm based on the identification of association rules. This is a generic approach, designed to put more stress on shallow linguistic processing, in order to deal in an efficient manner with real-world sentences. Domain ontologies are used to refine the set of relations, and the impact of this semantic filtering is evaluated through experiments conducted in the field of intelligence analysis.