An Improved Performance of Keyword Search Using Graph Structured Knowledge

Abstract

Keyword search is associate degree intuitive paradigm for looking joined knowledge sources on the online. We have a tendency to propose to route keywords solely to relevant sources to cut back the high value of process keyword search queries over all sources. We have a tendency to propose a unique methodology for computing top-k routing plans supported their potentials to contain results for a given keyword question. We have a tendency to use a keyword-element relationship outline that succinctly represents relationships between keywords and therefore the knowledge parts mentioning them. A structure evaluation mechanism is projected for computing the relevancy of routing plans supported scores at the amount of keywords, knowledge parts, component sets, and sub graphs that connect these parts. Experiments administrated mistreatment a hundred and fifty publically accessible sources on the online showed that valid plans (precision@1 of zero.92) that square measure extremely relevant (mean reciprocal rank of zero.89) are often computed in one second on the average on one laptop. Further, we have a tendency to show routing greatly helps to enhance the performance of keyword search, while not compromising its result quality.