Extracting Non-taxonomic Relationships ofOntologies from Texts

摘要

Manual construction of ontologies by domain experts and knowledgeengineers is an expensive and time consuming task so, automatic and/or semi-automatic approaches are needed. Ontology learning looks for identifyingontology elements like non-taxonomic relationships from information sources.These relationships correspond to slots in a frame-based ontology. This articleproposes an initial process for semi-automatic extraction of non-taxonomicrelationships of ontologies from textual sources. It uses Natural LanguageProcessing (NLP) techniques to identify good candidates of non-taxonomicrelationships and a data mining technique to suggest their possible best level in theontology hierarchy. Once the extraction of these relationships is essentially aretrieval task, the metrics of this field like recall, precision and f-measure are usedto perform evaluation.