Performance Evaluation of Desktop Search Engines

摘要

With the rapid increase in computer hard drive capacity, the amount of information stored on personal computers as digital photos, text files, and multimedia has increased significantly. It has become time consuming to search for a particular file in the sea of files on hard drives. This has led to the development of several desktop search engines that help locate files on a desktop effectively. In this paper, the performance of five desktop search engines, Yahoo, Copernic, Archivarius, Google, and Windows are evaluated. An established dataset, TREC 2004 Robust track, and a set of files representing a typical desktop have been used to perform comprehension experiments. A standard set of evaluation measures including recall-precision averages, document level precision and recall, and exact precision and recall over retrieved set are used. The evaluations performed by a standard evaluation program provide an exhaustive performance comparison of the desktop search engines by representative information retrieval measures.