RDF與Topic Maps之知識表徵比較研究
No Thumbnail Available
Date
2005
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
網路資源組織近來一直是圖書館界關心的課題之一，但一般來說，圖書館整理的資源向來是以實體為主，具有同質性高、定義嚴謹及穩定等特性，而網路資源相較於圖書館的資源，則具有大量、異質、分散、內容與位址經常改變、且成長快速等特性。不管是圖書館的索引典與詮釋資料架構（schema），或人工智慧領域過去已發展的一些知識表徵語言，例如CycL、Loom、與KIF等，都不見得可用來處理網路環境之半結構式資源。換言之，網路知識表徵需將網路資源的特性納入考量，重新思考與發展合適的處理方法。
目前「全球資訊網協會」（World Wide Web Consortium，簡稱W3C）及「國際標準組織」（International Organization for Standardization，簡稱ISO）在2000年前後，分別提出「資源描述架構」（Resource Description Framework，簡稱RDF）與「主題地圖」（Topic Maps），作為可應用於網路資源組織與知識表徵方法。RDF與Topic Maps都在表徵人類的知識關聯，雖均可為知識本體（Ontology）提供語意互通的資料模型（data model），但這兩種語言最初發展的目的並不相同，亦有其有各自的交換語法（interchange syntax），因此這兩種語言在知識關聯程度的表徵上到底有何不同、孰強孰弱、及適用範圍為何等，都是本研究想探討的課題。具體而言，本研究試圖回答三個問題：
一、網路知識表徵之發展趨勢為何？
二、RDF與Topic Maps在表達概念或主題關聯之語法、所能表達的語意、及應用情境等為何？異同為何？
三、RDF與Topic Maps在圖書館網路資源組織應用的可行性為何？
本研究對網路知識表徵之發展趨勢的瞭解是由文獻分析獲得；對RDF與Topic Maps在語法及所能表達的語意之異同，及圖書館網路資源組織的可行性，是藉由對語法的比較，及實際以RDF/XML與XTM對國家圖書館建置的「網路資源選介網站」進行編碼，就呈現結果檢視其異同等方式予以分析；應用情境則是以網路上蒐集到以RDF與Topic Maps實作的網站或系統來從事觀察。
研究結果主要分三方面，第一部分是有關網路知識表徵之發展趨勢，就巨觀層次而言，網路知識表徵之發展趨勢，在提供有助於人們建立與分享知識，並能協助機器對資源的自動處理，以提昇人們對資訊處理效率的網路環境。第二部分則是有關RDF與Topic Maps在知識表徵能力之比較，結果發現其各自在元素、語法、語意表達、與應用情境等面向不盡相同；此外，在協助資源瀏覽、資訊檢索、資訊過濾、及資源整合等能力亦各有所長。第三部分則嘗試探討RDF與Topic Maps在圖書館網路資源組織應用的可行性，依據實作經驗及呈現結果，認為RDF與Topic Maps在圖書館網路資源組織之應用具可行性，因利用其標記的網站具有分類架構較具彈性與延伸性、可提供使用者由不同面向瀏覽資源，以及有助於分類架構的交換互通、資源整合、與機器的自動處理等原「網路資源選介網站」所缺乏的特性。
最後，本研究依據研究結果，分別對RDF與Topic Maps在語法、應用、與實作方面提出建議。語法方面，建議限定RDF/XML語法之撰寫方式；對XTM則建議以URI作為指定「主題識別」的方式、修訂關聯不具方向的問題、以及改採支援XML Schema。應用方面，若需同時處理資源描述與主題索引的話，本研究建議兩種作法：一是對資源描述用RDF/XML標記，而對主題之索引用XTM標記，然後另外建立主題與資源的連結；二是完全採用RDF/XML來標記，而展現知識樹的方式就是在RDF/XML中建立表示主題的類，並利用屬性來表現類的階層與相關關係，甚至進一步促進社群內的成員，在表示相同的資源屬性與主題時，都使用相同的描述詞彙（例如Dublin Core）及相同的主題（例如rdf:ID指向相同的URI），以協助資源分享與整合。實作方面，建議使用廣為人知的詞彙，及建立或採用合適的Ontology。
The Internet resources organization has recently become one of the most important topics in Library and Information Science fields. Generally speaking, resources organized by libraries are presented in physical forms. They are similar in nature, well-defined, and relatively stable. In contrast, resources available on the Internet are massive, heterogeneous, distributed, and proliferative. Besides, their contents and site address change frequently. Both the thesauri and metadata built by libraries and the knowledge representation languages developed by artificial intelligence fields such as CycL, Loom, and KIF cannot be used directly to process semi-structured resources found on the Web. In other words, when generating a Web knowledge representation, all characteristics of Internet resources must be taken into careful consideration before coming up with a feasible solution. In order to organize Internet Resources and to represent Web knowledge, World Wide Web Consortium (W3C) and International Organization for Standardization (ISO) respectively proposed the Resource Description Framework (RDF) and the Topic Maps in 2000. Both RDF and Topic Maps were developed to manifest knowledge association of humans and were able to provide a data model of semantic interoperability for ontology. However, the original purposes of these two languages were not identical; each has its own interchange syntax. Therefore, the author believed that it’s necessary to find out the differences in the representation of knowledge association, the strengths and weaknesses, and the application feasibility between RDF and Topic Maps. Specifically, this study tries to answer the following research questions: 1. What are the trends in the development of the Web knowledge representation? 2. What are the similarities and differences in syntax for expressing the association of resources or topics between RDF and Topic Maps? What are the semantics these two languages can describe? To what extent can they apply? 3. When organizing Internet Resources, are RDF and Topic Maps feasible in libraries? The author observed trends in the development of Web knowledge representation through literature review. By comparing syntax and analyzing the outcomes of practically applying RDF/XML and XTM in National Central Library’s “Selected Internet Resources Website”, the author explored several similarities and differences in syntax, semantics, and the feasibility of Internet resources organization between RDF and Topic Maps. The results of this study are divided into three parts. The first part concerns the trends in the development of the Web knowledge representation. On a macro scale, the author finds the trend of establishing and sharing knowledge among humans and the trend of helping machines manage resources automatically, thereby enhancing the effectiveness of humans handling resources available on the Web. The second part is the comparison of knowledge representation capability between RDF and Topic Maps. The author discovered that there are differences in element, syntax, semantic representation, and application between RDF and Topic Maps. In addition, each one has its own advantages in helping browsing, information retrieval, information filtering, and resource integration. The third part is about the feasibility of the application of RDF and Topic Maps in libraries. The experiments show that Websites written by RDF/XML and XTM are flexible and extensible in classification structure. Besides, the resources in theses Websites can be browsed by different facets. Their interoperability of classification structures, resources integration, and automated processing capabilities of machines are better than Websites not written by these languages. Finally, this study offers suggestions for RDF and Topic Maps in syntax, application, and their practical application. On syntax part, it is suggested that the written format of RDF/XML should be limited. There are three suggestions for XTM. First, URI is recommended to serve as a reference for subject identity. Second, XTM should support XML Schema. Finally, the direction of association between topics should be restricted. On application part, the author suggests two ways that can be used to describe resources and index subjects at the same time. One is to use RDF/XML in resource description and to use XTM in subject index, then to build links between subjects and resources. The other is to use RDF/XML completely. In RDF/XML, subject index can be built by establishing classes of subjects and using properties to indicate the hierarchical relationships and correlations between classes. Furthermore, the author encourage members of a community to use identical vocabularies (e.g. Dublin Core) and topics (e.g. “rdf:ID” reference to the same URI) to improve the sharing and integration of resources when having the same resource properties and topics. In practical application, the use of popular vocabularies and appropriate ontology is recommended.
The Internet resources organization has recently become one of the most important topics in Library and Information Science fields. Generally speaking, resources organized by libraries are presented in physical forms. They are similar in nature, well-defined, and relatively stable. In contrast, resources available on the Internet are massive, heterogeneous, distributed, and proliferative. Besides, their contents and site address change frequently. Both the thesauri and metadata built by libraries and the knowledge representation languages developed by artificial intelligence fields such as CycL, Loom, and KIF cannot be used directly to process semi-structured resources found on the Web. In other words, when generating a Web knowledge representation, all characteristics of Internet resources must be taken into careful consideration before coming up with a feasible solution. In order to organize Internet Resources and to represent Web knowledge, World Wide Web Consortium (W3C) and International Organization for Standardization (ISO) respectively proposed the Resource Description Framework (RDF) and the Topic Maps in 2000. Both RDF and Topic Maps were developed to manifest knowledge association of humans and were able to provide a data model of semantic interoperability for ontology. However, the original purposes of these two languages were not identical; each has its own interchange syntax. Therefore, the author believed that it’s necessary to find out the differences in the representation of knowledge association, the strengths and weaknesses, and the application feasibility between RDF and Topic Maps. Specifically, this study tries to answer the following research questions: 1. What are the trends in the development of the Web knowledge representation? 2. What are the similarities and differences in syntax for expressing the association of resources or topics between RDF and Topic Maps? What are the semantics these two languages can describe? To what extent can they apply? 3. When organizing Internet Resources, are RDF and Topic Maps feasible in libraries? The author observed trends in the development of Web knowledge representation through literature review. By comparing syntax and analyzing the outcomes of practically applying RDF/XML and XTM in National Central Library’s “Selected Internet Resources Website”, the author explored several similarities and differences in syntax, semantics, and the feasibility of Internet resources organization between RDF and Topic Maps. The results of this study are divided into three parts. The first part concerns the trends in the development of the Web knowledge representation. On a macro scale, the author finds the trend of establishing and sharing knowledge among humans and the trend of helping machines manage resources automatically, thereby enhancing the effectiveness of humans handling resources available on the Web. The second part is the comparison of knowledge representation capability between RDF and Topic Maps. The author discovered that there are differences in element, syntax, semantic representation, and application between RDF and Topic Maps. In addition, each one has its own advantages in helping browsing, information retrieval, information filtering, and resource integration. The third part is about the feasibility of the application of RDF and Topic Maps in libraries. The experiments show that Websites written by RDF/XML and XTM are flexible and extensible in classification structure. Besides, the resources in theses Websites can be browsed by different facets. Their interoperability of classification structures, resources integration, and automated processing capabilities of machines are better than Websites not written by these languages. Finally, this study offers suggestions for RDF and Topic Maps in syntax, application, and their practical application. On syntax part, it is suggested that the written format of RDF/XML should be limited. There are three suggestions for XTM. First, URI is recommended to serve as a reference for subject identity. Second, XTM should support XML Schema. Finally, the direction of association between topics should be restricted. On application part, the author suggests two ways that can be used to describe resources and index subjects at the same time. One is to use RDF/XML in resource description and to use XTM in subject index, then to build links between subjects and resources. The other is to use RDF/XML completely. In RDF/XML, subject index can be built by establishing classes of subjects and using properties to indicate the hierarchical relationships and correlations between classes. Furthermore, the author encourage members of a community to use identical vocabularies (e.g. Dublin Core) and topics (e.g. “rdf:ID” reference to the same URI) to improve the sharing and integration of resources when having the same resource properties and topics. In practical application, the use of popular vocabularies and appropriate ontology is recommended.
Description
Keywords
知識表徵, 資源描述架構, 主題地圖, knowledge representation, Resource Description Framework, RDF, RDF/XML, Topic Maps, XTM