Acquisition of domain concepts and ontology construction
Qin Lu, Gaoying Cui · 2010
Ontology is an area of research that gets off to a flying start and attracts increasing attention in many branches of computer science as well as other relevant subjects. As a knowledge structure, ontology can be applied to different areas where domain knowledge is needed. Avoiding the weak points of manual methods of constructing ontologies, such as time consuming and difficult to update, automatic or semi-automatic ontology construction methods usually start with the acquisition of domain concepts from a proper corpus and then build the relations among these concepts. In this research, a series of studies have been conducted to acquire domain concepts and further build an ontology. The whole work can be divided into the following four aspects. The first is the acquisition of an appropriate domain corpus. Wikipedia, the world's largest online encyclopedia, is exploited for the acquisition of the corpus. Second, domain terms that are labels of concepts are extracted and given preliminary classification information, such as Part-of-Speech tags. A core term lexicon of domain can then be generated from the extracted terms, which are the most fundamental and frequently used terms in the domain and which comply with some specific standards as qualified core terms. Attributes of concept terms are also important components of domain concepts, because relations between concepts in an ontology are determined according to the relations between the attribute sets of these concept terms. Third, concept attributes and values are therefore defined, selected, and acquired from the given corpus according to syntactic information and also to instance information about the concepts. Last, with the help of annotated semantic information, concepts can be mined according to their corresponding instances. Based on all of the elements and relevant information extracted from the domain corpus, concepts filled with content are acquired and relations between them can be built to further construct a domain ontology accordingly.