Cross linguistic name matching in English and Arabic

Abstract

This paper presents a solution to the problem of matching personal names in English to the same names represented in Arabic script. Standard string comparison measures perform poorly on this task due to varying transliteration conventions in both languages and the fact that Arabic script does not usually represent short vowels. Significant improvement is achieved by augmenting the classic Levenshtein edit-distance algorithm with character equivalency classes.

Automatic Transliteration of Romanized Dialectal Arabic

“…Al-Onaizan and Knight (2002) introduced an approach for machine transliteration of Arabic names. Freeman et al (2006) also introduced a system for name matching between English and Arabic, which Habash (2008) employed as part of generating English transliterations from Arabic words in the context of machine translation. This work is similar to ours in terms of text transliteration.…”

Robust Dictionary Lookup in Multiple Noisy Orthographies

“…Thus, we find results that have low edit distance to our query. This approach is widely used in applications such as conventional spell correction, transliteration similarity and music information retrieval (Freeman et al, 2006;Toussaint and Oh, 2016).…”

“…Then, input terms are segmented into the available n-grams, and all possible transliterations are produced and scored based on their joint probabilities. Habash (2009) used an ambiguous mapping that utilized the soundslike indexing system Double Metaphones (Philips, 2000) combined with the direct mapping scores defined by Freeman et al (2006) to handle out-ofvocabulary words in the context of Arabic-English machine translation. Freeman et al (2006) extended Levenshtein Edit Distance to allow for improved matching of Arabic and English versions of the same proper names.…”

“…Habash (2009) used an ambiguous mapping that utilized the soundslike indexing system Double Metaphones (Philips, 2000) combined with the direct mapping scores defined by Freeman et al (2006) to handle out-ofvocabulary words in the context of Arabic-English machine translation. Freeman et al (2006) extended Levenshtein Edit Distance to allow for improved matching of Arabic and English versions of the same proper names. El-Kahky et al (2011) use graph reinforcement models to learn mapping between characters in different scripts in the context of transliteration mining.…”

Learning to match names across languages

“…Chinese ( normalized to a similarity score as in (Freeman et al 2006), where the score ranges from 0 to 1, with 1 being a perfect match. This edit-distance score is shown in the LEV row.…”