DefinePK

Abstract

Most of the present patent retrieval systems only use a single retrieval engine and cannot meet the growing demand of patent uses. Fusion of the results set of various retrieval engines is one of the best solutions for patent retrieval. In this study, we propose a Result Set Relevance Inference (RSRI) model for patent retrieval fusion. It makes use of two result sets returned from keyword retrieval and semantic retrieval engine and then fuses them to a new result set after calculation as per their weights. Using an experimental data set, results showed that the RSRI model can improve the recall and mean average precision in most cases.