We present a novel method for building a large-scale Japanese ontology from Wikipedia using one of the largest Japanese thesauri, Nihongo Goi-Taikei (referred to hereafter as 霨oi-Taikei? as an upper ontology. First, The leaf categories in the Goi-Taikei hierarchy are semi-automatically aligned with semantically equivalent Wikipedia categories. Then, their subcategories are created automatically by detecting is-a links in the Wikipedia category network below the junction using the knowledge deonned in Goi-Taikei above the junction. The resulting ontology has a well-deonned taxonomy in the upper level and a onne-grained taxonomy in the lower level with a large number of up-to-date instances. A sample evaluation shows that the precisions of the extracted categories and instances are 92.8% and 98.6%, respectively.
展开▼