Effective early termination techniques for text similarity join operator
Please cite this item using this persistent URLhttp://hdl.handle.net/11693/27360
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
- Conference Paper 
Text similarity join operator joins two relations if their join attributes are textually similar to each other, and it has a variety of application domains including integration and querying of data from heterogeneous resources; cleansing of data; and mining of data. Although, the text similarity join operator is widely used, its processing is expensive due to the huge number of similarity computations performed. In this paper, we incorporate some short cut evaluation techniques from the Information Retrieval domain, namely Harman, quit, continue, and maximal similarity filter heuristics, into the previously proposed text similarity join algorithms to reduce the amount of similarity computations needed during the join operation. We experimentally evaluate the original and the heuristic based similarity join algorithms using real data obtained from the DBLP Bibliography database, and observe performance improvements with continue and maximal similarity filter heuristics. © Springer-Verlag Berlin Heidelberg 2005.