Question answering towards automatic augmentations of ontology instances

Abstract

Ontology instances are typically stored as triples which associate two named entities with a pre-defined relational description. Sometimes such triples can be incomplete in that one entity is known but the other entity is missing. The automatic discovery of the missing values is closely related to relation extraction systems that extract binary relations between two identified entities. Relation extraction systems rely on the availability of accurately named entities in that mislabelled entities can decrease the number of relations correctly identified. Although recent results demonstrate over 80% accuracy for recognising named entities, when input texts have less consistent patterns, the performance decreases rapidly. This paper presents OntotripleQA which is the application of question-answering techniques to relation extraction in order to reduce the reliance on the named entities and take into account other assessments when evaluating potential relations. Not only does this increase the number of relations extracted, but it also improves the accuracy of extracting relations by considering features which are not extractable with only comparisons of the named entities. A small dataset was collected to test the proposed approach and the experiment demonstrates that it is effective on sentences from Web documents with an accuracy of 68% on average. © Springer-Verlag Berlin Heidelberg 2004.