資訊管理學報
戚玉樑;蔡明宏;
頁: 47-66
日期: 2007/07
摘要: 對建置本體(Ontology)於知識應用的開發者而言，定義合宜的本體概念是棘手且耗時的工作，目前大多數的實務領域均缺乏可供知識系統直接利用的概念定義，因此在真實世界的認知與系統所處理的知識之間，存有表達上的落差。由於文件為表達人類認知的重要媒介，因此已成為收集知識概念的主要來源之一，但文件原僅為提供人類形而上的理解，對於發展為系統可用的知識概念，則有下列問題須解決：1)．如何篩選出有意義的詞彙；2)．如何將詞彙發展為命名的概念；3)．如何建立概念之問的關係架構。本研究探討如何由文件中發展一套概念萃取的分析程序，它包含了文字探勘、語言學、心理及資料分析等學理方法，依據問題的特性及可行的解決方案整合成為可實際執行的“概念萃取程序”，最終將產出本體的概念雛型架構，以提供本體建置者的藍圖，並進一步發展為正式的領域本體。由實證的評估顯示，本程序可達到70%以上的主觀接受率，因此可做為建置者在開發本體概念架構初期的重要參考依據。
關鍵字: 本體;知識擷取;概念萃取;概念架構;
Abstract: For ontology developers, constructing comprehensible concepts of a know ledge domain is often problematic and time-consuming. The development process can be improved by reusing concept representation. However, the availability of reusable ontology is limited in practice due to the communication gaps existing between human cognition and machine-readable semantic representation required in know ledge engineering. Nowadays, textual documents that act as one of the most important medium to express human cognition are major resources for the acquisition of know ledge in computing systems. This study aims to develop a systematical approach to building reusable ontology prototypes from the documents. Several issues are identified such as how to recognize terminologies in textual corpus, to name concept tags in the terminologies, and to derive conceptual hierarchies. The proposed approach integrates concept extraction techniques that employ text-mining, linguistic, and psychological analysis methods. The empirical assessment reports the usability ratios of both conceptual tags and hierarchies exceed 70 percents. Consequently, the synergy of the elicitation techniques allows the developers to efficiently create and distribute ontology repository in specific domains.
Keywords: Ontology;Knowledge Acquisition;Concept Extraction;Conceptual Structure;
瀏覽次數: 15110 下載次數: 485
引用 導入Endnote
頁: 47-66
日期: 2007/07
摘要: 對建置本體(Ontology)於知識應用的開發者而言，定義合宜的本體概念是棘手且耗時的工作，目前大多數的實務領域均缺乏可供知識系統直接利用的概念定義，因此在真實世界的認知與系統所處理的知識之間，存有表達上的落差。由於文件為表達人類認知的重要媒介，因此已成為收集知識概念的主要來源之一，但文件原僅為提供人類形而上的理解，對於發展為系統可用的知識概念，則有下列問題須解決：1)．如何篩選出有意義的詞彙；2)．如何將詞彙發展為命名的概念；3)．如何建立概念之問的關係架構。本研究探討如何由文件中發展一套概念萃取的分析程序，它包含了文字探勘、語言學、心理及資料分析等學理方法，依據問題的特性及可行的解決方案整合成為可實際執行的“概念萃取程序”，最終將產出本體的概念雛型架構，以提供本體建置者的藍圖，並進一步發展為正式的領域本體。由實證的評估顯示，本程序可達到70%以上的主觀接受率，因此可做為建置者在開發本體概念架構初期的重要參考依據。
關鍵字: 本體;知識擷取;概念萃取;概念架構;
Knowledge Acquisition Approaches for Building Ontological Conceptual Prototypes in Document
Abstract: For ontology developers, constructing comprehensible concepts of a know ledge domain is often problematic and time-consuming. The development process can be improved by reusing concept representation. However, the availability of reusable ontology is limited in practice due to the communication gaps existing between human cognition and machine-readable semantic representation required in know ledge engineering. Nowadays, textual documents that act as one of the most important medium to express human cognition are major resources for the acquisition of know ledge in computing systems. This study aims to develop a systematical approach to building reusable ontology prototypes from the documents. Several issues are identified such as how to recognize terminologies in textual corpus, to name concept tags in the terminologies, and to derive conceptual hierarchies. The proposed approach integrates concept extraction techniques that employ text-mining, linguistic, and psychological analysis methods. The empirical assessment reports the usability ratios of both conceptual tags and hierarchies exceed 70 percents. Consequently, the synergy of the elicitation techniques allows the developers to efficiently create and distribute ontology repository in specific domains.
Keywords: Ontology;Knowledge Acquisition;Concept Extraction;Conceptual Structure;
瀏覽次數: 15110 下載次數: 485
引用 導入Endnote