Metallic materials ontology population from LOD based on conditional random field

• Published in: Computers in industry Vol. 99; pp. 140 - 155
• Main Authors: Zhang, Xiaoming, Zhang, Zhishen, Wang, Huiyong, Meng, Mingming, Pan, Dongyu
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.08.2018
• Subjects: Conditional random field (CRF); Linked open data; Metallic materials ontology; Ontology population; Ontology population; Metallic materials ontology; Linked open data; Conditional random field (CRF)
• DOI: 10.1016/j.compind.2018.03.032
• ISSN: 0166-3615

•This paper presents an approach to populate metallic materials ontology with LOD.•The filling position is obtained by the Conditional Random Fields algorithm.•This approach enriches the knowledge of metallic materials ontology.•Experiments show the feasibility and effectiveness of the method. In recent years, with the rapid development of ontology technology, many relatively perfect domain ontologies have emerged gradually and achieved favorable applications. However, for the existing metallic materials ontologies, such as the metallic materials ontology created by Ashino, MatonTO and ONTORULE, the knowledge of their instances is comparatively insufficient. Additionally, for the users, they hope that not only a large number of the materials instances are included in the ontology, but also the properties of the instances are desired. Linked Open Data (LOD) provides huge open knowledge bases which contain ample materials knowledge. Thus, we expect the knowledge of LOD can be inserted into a specific ontology. Obviously, it is not an easy work, since the LOD is very large, and its structure is inconsistent with ontology’s. Therefore, a method is proposed to populate a specific metallic materials ontology with the metallic materials information in the LOD. Firstly, in the LOD, we determine the information that can be filled into the existing metallic materials ontology. Then, we convert the LOD to Chain Triples (CHTs) according to the filling information. We use conditional random field (CRF) to achieve CHTs' filling positions in the specified metallic materials ontology. Finally, we insert the information into the ontology. The approach is evaluated in light of F-measure, and the experiment results demonstrate that the proposed approach can be effective to populate a specific ontology with the metallic materials data in LOD. This approach not only enriches the existing metallic materials ontology, but also greatly saves the manual efforts on the process of ontology population.