title: Similarity Search on the Web: Evaluation and Scalability Considerations
creator: Haveliwala, Taher
creator: Gionis, Aristides
creator: Klein, Dan
creator: Indyk, Piotr
subject: Data Mining
subject: Digital Libraries
description: Allowing users to find pages on the web similar to a particular query page is a crucial component of modern search engines.  A variety of techniques and approaches exist to support "Related Pages" queries.  In this paper we discuss shortcomings of previous approaches and present a unifying approach that puts special emphasis on the use of text, both within anchors and surrounding anchors.  In the central contribution of our paper, we present a novel technique for automating the evaluation process, allowing us to tune our parameters to maximize the quality of the results.  Finally, we show how to scale our approach to millions of web pages, using the established Locality-Sensitive-Hashing technique.
publisher: Stanford
date: 2001
type: Techreport
type: NonPeerReviewed
format: application/pdf
identifier: http://ilpubs.stanford.edu:8090/526/1/2001-8.pdf
identifier: Haveliwala, Taher and Gionis, Aristides and Klein, Dan and Indyk, Piotr (2001) Similarity Search on the Web: Evaluation and Scalability Considerations. Technical Report. Stanford.
relation: http://ilpubs.stanford.edu:8090/526/