Correlation Based Evaluation for Search Tools
Rajesh Kumar Goutam · 2018
An important task for information seeker is to find a search engine that requires minimum users efforts and delivers relevant and desired results in minimum span of time. It is likely the results to be organized in such a manner that relevancy of results decreases gradually as search length increases. In this paper, we propose a correlation based algorithm that accepts explicit users’ judgments about the relevancy of web documents in Ranked Precision Metric form. To remove the possibility of biased judgments, the algorithm incorporates implicit users feedback in the form of Session duration, Dwell time and click hits. We have conducted an extensive evaluation and comparison of three popular search engines with the help of 25 TREC queries. The paper presents the initial results derived with the help of Implicit users feedback and Ranked Precision metric.