Efficient indexing techniques for record matching and deduplication

Abstract

Record matching works on large sets of data, which may be either from single database or several databases. As size of database increases very rapidly, demand of matching process becomes too high. So, there is demand to minimise the number of matching pair records, time and cost in comparing records using efficient matching techniques. Recent researches have been done on record matching by number of researchers using various indexing techniques but as such they are not effective. Suffix array SA and q-gram are used indexing technique, but they lack somewhere in computation. This paper proposes two new indexing techniques: inverse suffix array ISA and Burrows-Wheeler transformation BWT to improve the performance of record matching process. The approach ISA can handle the multiple keywords simultaneously. We compare the performance of the proposed techniques with existing suffix array and q-gram indexing techniques and found that the new techniques are better than the earlier techniques.