Abstract:
Metadata has traditionally centered on descriptive representation through title, author, publisher, subject keywords and other attributes of scholarly output. The descriptive-centric metadata, however, has become increasingly inadequate as the complexity and volume of scholarly output grows. In this research, novel structural and referential metadata were proposed and created by utilizing sophisticated natural language processing, meta-search, and machine learning methodologies + crowdsourced user knowledge. The preliminary experiment and evaluation results show that the machine-induced with user-enhanced metadata can effectively facilitate scientific information retrieval, bibliometrics, publication ranking, summarization, and Cyberlearning.