Application of phonetic encoding for analyzing similarity of patient's data: Bangladesh perspective

Due to illiteracy and lack of standardized healthcare systems, patients in Bangladesh usually, provide misspelled names while making their entry via these systems. Different records with slightly misspelled names are thus generated which makes data mining and others tasks quite challenging and inefficient. In this paper, we have looked into the underlying problem of misspelled names of patients in healthcare systems and proposed a modified version of NameSignificance algorithm. Our proposed algorithm has performed significantly better than the existing solutions like NameSignificance, Modified Soundex, Double Metaphone encoding for Bangla as we founded our algorithm on the phonetic nature of Bengali names written in English transliterated form. Our algorithm achieved a staggering 77% matching of names whereas relevant algorithm could not pass 70% correct matches. This proposed method could pave the way for better record linkage and data analysis of the medical patient dataset. Our syllable based approach also helped us identify the reason behind the wrong matches appeared through our algorithm which will surely pave the way to purify our algorithm in the future.