An information Extraction Process forSemi-automatic Ontology Population

摘要

Most knowledge is available in unstructured texts, however, it must berepresented and handled automatically to become truly useful for the constructionknowledge-based systems. Ontologies are an approach for knowledge representa-tion capable of expressing a set of entities and their relationships, constraints, axi-oms and vocabulary of a given domain. Ontology population looks for identifyinginstances of concepts, relationships and properties of an ontology. Manual popula-tion by domain experts and knowledge engineers is an expensive and time con-suming task so, automatic or semi-automatic approaches are needed. This articleproposes a process for semi-automatic population of ontologies from text focusingon the application of natural language processing and information extraction tech-niques to acquire and classify ontology instances. Some experiments using a legalcorpus were conducted in order to evaluate it. Initial results are promising and in-dicate that our approach can extract instances with high effectiveness.