非監督式之本體論語意關聯擷取

Abstract

透過本體論模型，知識能更快速的分享與重複使用。除了用來表達知識外，本體論應用的範圍仍相當廣泛，常見如語意網、專家系統與資訊檢索系統等等領域都能看到本體論的蹤影。本體論能賦予資訊系統智能，有效提升舊有系統的效能；然而，建構本體論模型並非易事，經過詳盡地計畫、繁複地資料分析與專家協商，本體論模型方能產生。開發時所需的昂貴成本以及繁瑣步驟，造成本體論模型無法普及地用於解決實際問題。本研究中，基於現有的階層式知識架構，藉由階層關聯所提供的資訊，並透過大量文件，利用非監督式學習的方法，進一步擴充知識概念間的語意關聯，更快速地建構各組織所需之本體論模型，並減少開發所需成本。

With ontology, knowledge sharing and reusing can be faster and easier. The Ontology-based applications are various. Besides knowledge representation, ontology can be applied to semantic web, expert system, information retrieval system, and so on. Ontology provides intelligence for the information systems, and promotes the performance. Unfortunately, ontology development is a hard work that requires complicated tasks, including ontology planning, data analysis, and expert negotiation. The expensive cost of development leads to few ontology practice in the enterprises. In this paper, we propose an unsupervised approach for extracting semantic relations. We use text mining and the existent hierarchical knowledge structure to expand the semantic relations between the knowledge concept pairs. The proposed approach can increase the speed of ontology construction, and reduce human resource demands, furthermore, make ontology practice in the real world be possible.