A Practical Technique for Thai-English Word Mapping Using Phonetic Rules: Person Name Matching Case Study

This paper proposes a practical technique for matching person names between Thai and English based on phonetical rules to map the same name represented in different scripts. We applied the romanization approach to convert Thai names to Roman alphabets. Then we proposed vowel processing and consonant processing to reform these romanized names. Our proposed convert some sets of roman alphabets to a single alphabet based on phonetics. Then similarity measurements such as Levenshtein distance and Jaro-Winkler are used for the similarity measure between the romanized and English names. To evaluate the effectiveness of our proposed approach, we used a confusion matrix to compare its performance against that of the baseline experiment. Our proposed method with the Levenshtein distance metric proved to be the most effective, with an average accuracy of 99.98%, precision of 96.68%, recall of 98.24%, and F1-score of 97.45%. These studies show the proposed method can be used to enhance the effectiveness of similarity measures for name matching in Thai language datasets, especially using with the Levenshtein distance.