Swoosh： a generic approach to entity resolution教程.pdfVIP

The VLDB Journal (2009) 18:255–276 DOI 10.1007/s00778-008-0098-x REGULAR PAPER Swoosh: a generic approach to entity resolution Omar Benjelloun · Hector Garcia-Molina · David Menestrina · Qi Su · Steven Euijong Whang · Jennifer Widom Received: 22 May 2007 / Revised: 9 January 2008 / Accepted: 10 January 2008 / Published online: 11 March 2008 © Springer-Verlag 2008 Abstract We consider the entity resolution (ER) problem Keywords Entity resolution · Generic entity resolution · (also known as deduplication, or merge–purge), in which Data cleaning records determined to represent the same real-world entity are successively located and merged. We formalize the gen- eric ER problem, treating the functions for comparing and 1 Introduction merging records as black-boxes, which permits expressive and extensible ER solutions. We identify four important prop- Entity resolution (ER) (sometimes referred to as deduplica- erties that, if satisfied by the match and merge functions, tion) is the process of identifying and merging records judged enable much more efficient ER algorithms. We develop three to represent the same real-world entity. ER is a well-known efficient ER algorithms: G-Swoosh for the case where the problem that arises in many applications. For example, mail- four properties do not hold, and R-Swoosh and F-Swoosh that ing lists may contain multiple entries representing the same exploit the four properties. F-Swoosh in addition assumes physical address, but each record may be slightly different, knowledge of the “features” (e.g., attributes) used by the e.g., containing different spellings or missing some informa- match function. We experimentally evaluate the algorithms tion. As a second example, consider a company that has dif- using comparison shoppin