Abstract
With the aim of reducing duplicate records in databases, duplicate record detection (DRD) ensures the integrity of data. Its role is to identify records signifying same entities either in the same or in different compared to database. A diversity of indexing techniques has been proposed to support DRD. Q-gram is one of the common techniques used to index databases. This paper introduces modification to the Q-gram indexing technique. Such modification participates in improving the performance of the duplicate detection process and in reducing the time and number of comparisons. In the proposed work, in order to make the back-end computations easier, Q-gram strings are alternatively converted into numeric values using their corresponding ASCII code. Based on these numeric values, the indexing will decrease the complexity of Q-gram comparisons and speed up the DRD process as a whole. Unlike the existing approaches, the proposed technique is easier in implementation and requires less memory space. Two other variations of the proposed technique are introduced in this paper to decrease the matching process time; the first uses a range for matching, while the second sorts words alphabetically inside blocks. According to experimental results, the three proposed techniques perform much faster and are almost as accurate as the current Q-gram technique, meaning that they can be used in large-sized databases DRD.