ME-Match: Tonal Grouping Based Approach in Cross-Script Name Matching

Abstract

Even though matching between different scripts could be immensely useful for news organizations, author recognition with cross-script matches in digital libraries and homeland security, it is impossible to automatically match. Now, we propose a new approach, ME-Match, for matching the proper names across different scripts. The foremost concept of our approach is to match them via phoneme strings. The main steps in ME-Match are creation of bilingual pronouncing mapping, tokenization of query names, transformation of query names to IPA forms based on tonal grouping approach, searching possible various words in both scripts for each query IPA phoneme string, combination of various words to become full name strings and then searching names. The performance is measured by standard information-retrieval metrics: recall, precision, and f-measures.