Extended E-N-DIST Algorithm for Alias Detection

Abstract

Nowadays personal names are not the only way to refer to celebrities and experts from different fields, instead, they can be referred to by their aliases on the web. Associated aliases have remarkable importance in retrieving information about the personal name from the websites. Therefore, disclosing aliases can have an important role in overcoming many real-world challenges. In this research, the aim is to explore and propose a reliable algorithm that can detect aliases that occurred due to transliteration of Arabic names into English. An extension to the Enhanced N-gram distance algorithm (E-N-DIST) which was previously published is introduced in this paper. The proposed algorithm is called the Extended Enhanced N-gram distance algorithm (E-E-N-DIST). The differences between E-N-DIST and E-E-N-DIST are two main changes in calculating the cost of substitution and transposition. First, E-E-N-DIST is computed based on 2n+1 – 1 states. The second is the use of an edit operation called the ’Exchange of Vowels’ to count the common spelling errors that happen due to the transliteration from one language to another. The idea of exchange of vowels is to search for vowels (viz. = a‘, = e‘, = i’, = o‘, and = u‘) and the non-vowel character = y‘ that has a vowel sound or a part of it in other languages to estimate the operations cost of insertion and deletion. The proposed algorithm tested using a dataset for the literature; the results obtained are compared with other algorithms from the state of the art. The proposed algorithm outperforms other algorithms; it achieved a better average percentage of similarity than all other compared algorithms.

Keywords

• Alias detection
• edit distance (ED)
• Levenshtein distance (LD)
• E-N-DIST
• dynamic programming