Building ontologies is a difficult and time-consuming task, requiring to combine the knowledge of domain experts with the skill and experience of ontology engineers resulting in a high demand on scarce expert resources. Moreover, the size of knowledge bases needed in real-world applications easily exceeds the modeling capabilities of any human expert. On the other hand, both quality and expressivity of the ontologies generated automatically by the state-ofthe-art ontology learning systems fail to meet the expectations of people who argue in favor of powerful, knowledge-intensive applications based on ontological reasoning. In order to overcome this bottleneck, it is necessary to thoroughly assist the
modeling process by providing hybrid semi-automatic methods that (i) intelligently suggest the extraction of potential knowledge elements (complex domain axioms or facts) from resources such as domain relevant text corpora and (ii) provide guidance during the knowledge specification process by asking decisive questions in order to clarify still undefined parts of the knowledge base. Obviously, those two requirements complement each other. The first one
clearly falls into the area of natural language processing. By using existing methods for knowledge extraction from texts, passages can be identified that indicate the validity of certain pieces of knowledge. For the second requirement, strictly logic-based exploration techniques are needed that yield logically crisp propositions. We believe that integrating these two directions of knowledge acquisition in one scenario will help to overcome disadvantages of either approach. The framework proposed in this chapter realizes this integration and shows its potential for practical applications. In Section 9.2, we briefly introduce the description logic SHOIN . Sec-
tion 9.3 sketches the field of ontology learning before presenting LExO as one method for acquiring DL axioms from texts. Section 9.4 gives the necessary background for Relational Exploration (RE), a technique used for interactive knowledge specification based on Formal Concept Analysis. In Section 9.5, we describe in detail how LExO and RE (possibly assisted by other ontology learning components) can be synergetically combined into an integrated framework. Implementation details as well as an example are given in Section 9.6. Finally, Section 9.7 concludes and gives an outlook to future research.