Re-Ranking Web Data Per A Structured Domain
Grace Zhao · International Journal of Service and Knowledge Management · 2019
To build a domain-specific knowledge base of web resources, we have to face the challenge of having the computers gather domain-relevant knowledge feeds. We propose a re-ranking algorithm to effectively fetch and re-group the web data crawled by some credited web search engines, to meet the needs. The algorithm studies the structure and semantics of the domain ontology (graph) and constructs computational relations among nodes. After examining matching terms between ontology dictionary and the textual content (text, metadata) of the retrieved documents, we calculate three-dimensional information scores -- distance, direction, and relationship of each document in relation to the query string that corresponds to a node in the ontology graph. Further, we explore the directional relation and study three sub-degrees: granularity, diversity, and generality. Based on the multivariate information degrees, we subsequently re-rank the retrieved documents and provide more meaningful data to the domain space.