A Multiple-Domain Ontology Builder

Proceedings of the 23rd International Conference on Computational Linguistics (Coling 2010), pages 967–975, Beijing, August 2010 A Multiple-Domain Ontology Builder Sara Salem Samir AbdelRahman Abstract The interpretation of a multiple-domain text corpus as a single ontology leads to misconceptions. This is because some concepts may be syntactically equal; though, they are semantically lopsided in different domains. Also, the occurrences of a domain concept in a large multiple- domain corpus may not gauge correctly the concept significance. This paper tackles the mentioned problems and pro- poses a novel ontology builder to extract separate domain specific ontologies from such a corpus. The builder contribution is to sustain each domain specific concepts and relations to get precise answers for user questions. We extend a single ontol- ogy builder named Text2Onto to apply our thought. We fruitfully enhance it to answer, more precisely, questions on a subset of AQUAINT corpus. 1 Introduction Domain ontology is a knowledge representation of the domain as a set of concepts and relations. Ontology notion always presents handy semantic solutions for various hot research areas such as Semantic Web, Informational Retrieval, and Question Answering. Currently, automatic ontology builders pre- sume that the given corpus has a single domain. When used with a multiple-domain corpus, these builders generate 1 large ontology for the whole corpus. Dramatically, this causes 2 domain mis- conception problems. First, the ontology concep- tual model becomes imprecise for the common concepts in various domains having different semantics. Second, the relevance weights as- signed to the concepts do not measure precisely their significance in specific domains. This paper presents a promising solution for the mentioned problems. The proposed solution is an integrated 2-layer ontology builder. The ontology layers are: 1) the conceptual layer, which has the key concepts and relations of each separate domain, and 2) the general layer, which maintains the general domain information re- garding related persons, organizations, locations, and dates. Our proposed 2-layer ontology im- proves the extracted answers for single-domain and cross-domain questions. We successfully prove our thought against Text2Onto builder. Ontology extraction from a domain corpus has been targeted by many researchers. The core ex- traction approaches can be classified into 3 ap- proaches. The first approach is to build the on- tology from scratch (Buitelaar et al., 2004; Ci- miano and Völker, 2005). The second approach is to extend a predefined general ontology, such as WordNet, with possible application domain concepts and relations (Navigli and Velardi, 2004). The last approach is to build ontology as a composition of other predefined ontologies (Ci- miano et al., 2006). Moreover, as an ontology building design decision, the resultant ontology is either a single layer ontology or a multi- layered ontology (Benslimane et al., 2000; Du- montier and Villanueva-Rosales, 2007). The paper is organized as follows: Section 2 introduces some related systems; Section 3 ex- plains the misconceptions due to extracting a single ontology from a multiple-domain corpus; Section 4 describes our proposed builder; Section 5 illustrates our Question Answering system, which is used for …

Content truncated to 500 words