Identification of Similar Strings in a Dataset using Scalable Join

Abstract

Similarity Join plays an important role in data integration and cleansing, record linkage and data de-duplication. It finds similar sting pairs from collections of strings. If two strings are similar they share a common token. Number of approaches has been proposed for in-memory string similarity joins. But due to the rising era of big data, demands for scalable algorithms to support large scale string similarity joins arises. The proposed architecture uses the MapReduce concept and is based on inverted index and multiple prefix filtering methods. The prefix filtering is made of different global ordering which reduces the number of candidate pairs significantly, thus improving the pruning power as compared to other approach.