Log in / Register
Home arrow Computer Science arrow Linked Open Data
< Prev   CONTENTS   Next >

6.2 Interlinking Korean Resources in Korean and English: Korean Transliteration Distance

During translation some terms may be transliterated from one language to another in case translators cannot find proper corresponding terminology in that local language. Transliteration is a way of converting letters from one writing system into another without concern for representing the original phonemics. For example, a Korean popular food called (translated as knife-cut Korean noodles) can be transliterated as kalguksu in English.

The best approach to measure distance between transliterated strings would be to apply Korean Phoneme Distance to the strings transliterated back. This approach, however, is complicated because a transliterated Korean string could be back transliterated into several possible original Korean strings in case the transliterated string has no explicit notation for identifying syllables. Unfortunately, existing transliteration systems do not consider back-transliteration, so no explicit borders of syllables, which is important to restore the original Korean words, exist. Although many efforts have focused on the back transliteration to help people identify the original string better, many existing Korean-English transliteration systems lack this mechanism.

Due to this difficulty, we decided to take a simpler but more practical approach for measuring distance between transliterated Korean strings, namely, Korean Transliteration Distance, which chooses one random letter for each consonant group from the International Phonetic Alphabet (IPA) chart. For example, as 'b' and 'p' belong to the bilabial plosive group, Korean Transliteration Distance replaces every 'b' with 'p'. Similarly, it replaces 'g' with 'k', 'd' with 't', and 'l' with 'r' (although 'l' and 'r' do not belong to the same group, they are used interchangeably in Korea). The main difference between Soundex and Korean Transliteration Distance is that Korean Transliteration Distance does not eliminate vowels or other consonants, does not remove duplicates, and does not limit the number of letters for comparison. There are three reasons for this. First, the Korean alphabet has 10 vowels and 14 consonants compared with 5 vowels and 21 consonants in English, so Korean vowels play a more important role in matching words than English vowels do. Second, the Korean alphabet has letters with fortis, which is expressed with two identical consonants in succession, so the meaning of the string will be lost if duplicate consonants are removed. Third, keeping all the letters is a more conservative and safer way to measure distance. This metric is especially useful for enhancing recall while keeping precision almost intact compared with Levenshtein and for enhancing precision compared with Soundex, thereby contributing to obtaining a higher number of correct links when used in Silk.

Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >
Business & Finance
Computer Science
Language & Literature
Political science