Metadata Architecture for Digital Library Integration

• Main Authors: Su-Hsien Huang, 黃夙賢
• Other Authors: Wei-Pang Yang
• Format: Others
• Language: en_US
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/93894890887154828752

博士 === 國立交通大學 === 資訊科學與工程研究所 === 95 === Metadata has been playing an essential role in integrating heterogeneous digital libraries (DL). However, conventional metadata architecture is insufficient to achieve interoperability among DL because of the heterogeneity in semantics and no structure consideration in metadata formats. This dissertation proposes novel metadata architecture called M-Architecture@DL to integrate DL seamlessly from the perspective of metadata. M-Architecture@DL follows Model-Extraction-Query (MEQ) model to obtain more permanent and explicit knowledge in the process of DL query. M-Architecture@DL contains three layers, namely metadata modeling layer, data extraction layer, and semantic query layer. The separation of M-Architecture@DL into three-layer achieves format, protocol and semantic interoperability in each layer. Metadata modeling layer uses Metadata Modeling Language (MML) to describe real-world entities. MML adopts XML as its syntax and extends Resource Description Framework (RDF) by adding name hierarchy reference. MML provides two constructors, tuple and set constructors, to represent structures. With these two constructors, metadata can be translated by manipulating attributes of metadata with operations. In this layer, the format interoperability is achieved. Data extraction layer collects data from distributed DL and encapsulates result into MML metadata. Data from DL services with similar structure can be extracted into metadata automatically by means of the common structure. In the process of extraction, the first step is to assign level ID for the sample document and determine the common part to be extracted. Then an extraction algorithm called Metadata Extractor is implemented to extract the documents according to the common structure. This layer provides a transparent way without prearrangement with distributed DLs and saves much effort to collect information through the HTTP protocol. Therefore, the protocol interoperability is achieved. Semantic query layer retrieves metadata semantically by adding relationships in query statements. A Content and Service Inference Model (CSIM) is proposed to derive 15 relationships from two essential aspects of DL: content and services. The 15 structural relationships create operations to manipulate metadata in a query predicate and facilitate a query with as much semantics. Both content and service queries are presented to derive more semantic answers in a DL search. In this layer, the semantic interoperability is achieved. Experiments are conducted and indicate that M-Architecture@DL has excellent performance in DL integration. The experiment results have shown that both accuracy and coverage are improved to a conventional keyword-based approach. Adopting M-Architecture@DL can alleviate the administrative load. When developing novel DL services, such as library resource planning and virtual union catalog system, librarians are recommended with alternative answers to combine existent DL components. The reuse of DL services and metadata is the future trend in DL integration.