TR-IIS-05-013    Fulltext


Active Feedback for Effective Web Search

Ray-I Chang, Jan-Ming Ho

Abstract

Current search engines with arrays of servers can provide efficient web content services. However, as their returned results are usually enormous mass, finding target information is still time-consuming. To increase the correctness of returned results, different page ranking methods were introduced. Some of them also try to use users’ feedback to increase their precision in ranking. However, as the traditional approaches are passive in feedback and greedy in search, experiments show that the average error ratio is over 20%. Their returned results are usually too large to satisfy users’ needs. In this paper, an active feedback technology is introduced. It bases on the concept of balanced tree to present some critical questions for guiding users to have the proper feedback in further searching. The same idea can be applied to assist either distributed or P2P (peer-to-peer) search engines to balance workloads and speed responses.

Keywords: active feedback, balanced tree, distributed, peer-to-peer, search engine