- 學位論文
權重導向分析於本體論相似度計算之研究
A Study on Weight-Oriented Similarity Calculation for Ontology
摘要
隨著建置本體論的方法增加，使得本體論的建置更為便利，因而產生的許多的本體論，並且也出現了許多相同領域的本體論。但是建置本體論需要耗費大量人力與時間，因此若能夠將相同領域的兩個本體論互相合併，就能夠減少建置本體論所帶來的成本。一般合併本體論的方式，都是利用其他相同領域之本體論的內容來增加本身所使用之本體論的知識，但是通常都沒有考量到合併後的本體論是否還可以使用。此時就需要探討合併後的本體論，本身的結構是否扭曲或是加入不相關的概念等，造成應用上提高複雜度以及產生概念分散的問題。本論文透過WordNet本身的本體論來計算概念之間的屬性相似度和語意相似度，並藉由門檻值設定，將相似度較高的概念做合併，避免不相關的概念加入本體論。接著計算深度相關語意相似度以及概念間的關係，做為本體論統整的依據，進而提升本體論的可利用性。
並列摘要
With the increase of different ontology building approaches, ontologies have been produced in different ways even in the same domain. Ontology building requires a lot of effort and time, but we can reduce the building costs if the existing ontologies in the same domain can be used as the base for constructing a new ontology. A general ontology merging approach uses the contents of the target ontology to increase the knowledge of the domain ontology, but it does not take the applicability of ontologies into consideration and it may raise the complexity of the merging procedure. Therefore, our proposed method is to explore the structure of the merged ontology, checking if any irrelevant concept has been added. In this thesis, we calculate the concepts between feature matching and semantic matching through WordNet and merge the concept which has the highest similarity value while filtering irrelevant concept which scores below the threshold. Finally, we calculate the depth relative measures and relationships between concepts as a basis for ontology integration and furthermore enhance the usability of ontology.