សួស្តី (suosdei) to Khmer Romanization part II
Bài đăng này đã không được cập nhật trong 4 năm
សួស្តី (suosdei or suostei)
In my last article, we have implemented an algorithm for converting Khmer character to Roman character. The algorithm is very easy to implement where we just replace Khmer alphabets with English alphabets. By testing on few word, I though that my algorithm is ok and it's ready for production.
Eyes opening about Khmer Romanization
However, after I met Mr. Chem Vatho who proposed to me the idea of Khmer Romanization. During our meeting, he had taught and shown me about the complexity of Khmer linguistic especially abput Khmer Romanization. He had point out many flaws of my algorithm where it could not produce a correct Roman word for some conditions. Moreover, he drew flow of the complex rule of Khmer Romanization to me as a guide line for improving my algorithm. It's an eyes opening for me about the complexity of Khmer language.
Refactor the algorithm
After my meeting with Mr. Vatho, now we can improve our algorithm here.
In the file consonant.go, we use it for checking Khmer character to is it in group អ or អ៊ since each group of consonant can affect to Raman character of next vowel.
//consonant.go
package romanization
func IsKorGroup(ch rune) bool {
korList := map[rune]string{
6018: "k", //គ
6019: "kh", //ឃ
6020: "ng", //ង
6023: "ch", //ជ
6024: "chh", //ឈ
6025: "nh", //ញ
6028: "d", //ឌ
6029: "th", //ឍ
6033: "t", //ទ
6034: "th", //ធ
6035: "n", //ន
6038: "p", //ព
6039: "ph", //ភ
6040: "m", //ម
6041: "y", //យ
6042: "r", //រ
6043: "l", //ល
6044: "v", //វ
}
_, isKorWork := korList[ch]
return isKorWork
}
func IsConsonant(ch rune) bool {
consonantList := map[rune]string{
6016: "k", //ក
6017: "kh", //ខ
6018: "k", //គ
6019: "kh", //ឃ
6020: "ng", //ង
6021: "ch", //ច
6022: "chh", //ឆ
6023: "ch", //ជ
6024: "chh", //ឈ
6025: "nh", //ញ
6026: "d", //ដ
6027: "th", //ឋ
6028: "d", //ឌ
6029: "th", //ឍ
6030: "n", //ណ
6031: "t", //ត
6032: "th", //ថ
6033: "t", //ទ
6034: "th", //ធ
6035: "n", //ន
6036: "b", //ប
6037: "ph", //ផ
6038: "p", //ព
6039: "ph", //ភ
6040: "m", //ម
6041: "y", //យ
6042: "r", //រ
6043: "l", //ល
6044: "v", //វ
6045: "s", //ឝ
6046: "s", //ឞ
6047: "s", //ស
6048: "h", //ហ
6049: "l", //ឡ
6050: "a", //អ
}
_, isConsonant := consonantList[ch]
return isConsonant
}
func Roman(ch rune) string {
wordsList := map[rune]string{
6016: "k", //ក
6017: "kh", //ខ
6018: "k", //គ
6019: "kh", //ឃ
6020: "ng", //ង
6021: "ch", //ច
6022: "chh", //ឆ
6023: "ch", //ជ
6024: "chh", //ឈ
6025: "nh", //ញ
6026: "d", //ដ
6027: "th", //ឋ
6028: "d", //ឌ
6029: "th", //ឍ
6030: "n", //ណ
6031: "t", //ត
6032: "th", //ថ
6033: "t", //ទ
6034: "th", //ធ
6035: "n", //ន
6036: "b", //ប
6037: "ph", //ផ
6038: "p", //ព
6039: "ph", //ភ
6040: "m", //ម
6041: "y", //យ
6042: "r", //រ
6043: "l", //ល
6044: "v", //វ
6045: "s", //ឝ
6046: "s", //ឞ
6047: "s", //ស
6048: "h", //ហ
6049: "l", //ឡ
6050: "a", //អ
6051: "a", //ឣ
6052: "a", //ឤ
6053: "I", //ឥ
6054: "I", //ឦ
6055: "U", //ឧ
6056: "Uu", //ឨ
6057: "Uu", //ឩ
6058: "Au", //ឪ
6059: "Ry", //ឫ
6060: "Ryy", //ឬ
6061: "Ly", //ឭ
6062: "Lyy", //ឮ
6063: "E", //ឯ
6064: "Ai", //ឰ
6065: "Ao", //ឱ
6066: "Ao", //ឲ
6067: "Au", //ឳ
6100: ".", //។
6101: ".", //៕
6102: ":", //៖
6104: ".etc.", //៘
6128: "0", //៰
6129: "1", //៱
6130: "2", //៲
6131: "3", //៳
6132: "4", //៴
6133: "5", //៵
6134: "6", //៶
6135: "7", //៷
6136: "8", //៸
6137: "9", //៹
6112: "0", //០
6113: "1", //១
6114: "2", //២
6115: "3", //៣
6116: "4", //៤
6117: "5", //៥
6118: "6", //៦
6119: "7", //៧
6120: "8", //៨
6121: "9", //៩
}
if r_ch, isMatch := wordsList[ch]; isMatch {
return r_ch
}
return ""
}
In the file vowel.go, we use it for checking if the before Khmer character to is it in group អ or អ៊, so it could return the correct Raman alphabet. Moreover, we also implement logic to check for some special cases of Khmer vowel.
//vowel.go
package romanization
var vowelForK = map[rune]string{
6070: "a", //ា
6071: "e", //ិ
6072: "ei", //ី
6073: "oe", //ឹ
6074: "eu", //ឺ
6075: "o", //ុ
6076: "ou", //ូ
6077: "uo", //ួ
6078: "aeu", //ើ
6079: "oea", //ឿ
6080: "ie", //ៀ
6081: "e", //េ
6082: "ae", //ែ
6083: "ai", //ៃ
6084: "ao", //ោ
6085: "au", //ៅ
6086: "am", //ំ
6087: "ah", //ះ
6088: "ak", //◌ៈ
6089: "", //=៉
6091: "", //់
6092: "r", //៌
6095: "a", //៏
6096: "oa", // ័
6098: "", //=្
}
var vowelForKor = map[rune]string{
6070: "ea", //ា
6071: "i", //ិ
6072: "i", //ី
6073: "ue", //ឹ
6074: "ueu", //ឺ
6075: "u", //ុ
6076: "u", //ូ
6077: "uo", //ួ
6078: "au", //ើ
6079: "oea", //ឿ
6080: "ie", //ៀ
6081: "e", //េ
6082: "eae", //ែ
6083: "ey", //ៃ
6084: "ou", //ោ
6085: "ov", //ៅ
6086: "um", //ំ
6087: "eah", //ះ
6088: "eak", //◌ៈ
6089: "", //=៉
6091: "", //់
6092: "r", //៌
6095: "a", //៏
6096: "oa", // ័
6098: "", //=្
}
func IsVowel(ch rune) bool {
_, isVowel := vowelForK[ch]
return isVowel
}
func IsSuffixableVowel(ch rune) bool {
suffixableVowelList := map[rune]string{
6070: "ea", //ា => ាំ
6071: "i", //ិ => ិះ
6075: "u", //ុ => ុះ
6081: "e", //េ => េះ
6084: "ou", //ោ => ោះ
6086: "am", //ំ => ំុ
6096: "oa", // ័
}
_, isSpecial := suffixableVowelList[ch]
return isSpecial
}
func SpecialVowel(first rune, second rune, isKor bool) string {
var specialVowelList map[rune]string
if isKor {
specialVowelList = map[rune]string{
12161: "um", //ំ => ំុ, 6086 + 6075 = 12161
12156: "oam", //ា => ាំ, 6070 + 6086 = 12156
18176: "eang", //ាំង, 6070 + 6086 + 6020= 18176
12168: "eh", //េ => េះ, 6081 + 6087 = 12168
12162: "uh", //ុ => ុះ, 6075 + 6087 = 12162
12171: "uoh", //ោ => ោះ, 6084 + 6087 = 12171
12158: "is", //ិ => ិះ, 6071 + 6087 = 12158
12116: "eang", //័ង , 6096 + 6020 = 12116
12112: "eak", //័ក , 6096 + 6016 = 12112
12137: "ey", //័យ , 6096 + 6041 = 12137
}
} else {
specialVowelList = map[rune]string{
12161: "om", //ំ => ំុ, 6086 + 6075 = 12161
12156: "am", //ា => ាំ, 6070 + 6086 = 12156
18176: "ang", //ាំង, 6070 + 6086 + 6020= 18176
12168: "eh", //េ => េះ, 6081 + 6087 = 12168
12162: "oh", //ុ => ុះ, 6075 + 6087 = 12162
12171: "aoh", //ោ => ោះ, 6084 + 6087 = 12171
12158: "eh", //ិ => ិះ, 6071 + 6087 = 12158
12116: "ang", //័ង , 6096 + 6020 = 12116
12112: "ak", //័ក , 6096 + 6016 = 12112
12137: "ai", //័យ , 6096 + 6041 = 12137
}
}
if r_ch, isMatch := specialVowelList[first+second]; isMatch {
return r_ch
}
if IsConsonant(second) {
return VowelWord(first, isKor) + Roman(second)
}
return VowelWord(first, isKor) + VowelWord(second, false)
}
func VowelWord(ch rune, isKor bool) string {
var wordsList map[rune]string
if isKor {
wordsList = vowelForKor
} else {
wordsList = vowelForK
}
if r_ch, isMatch := wordsList[ch]; isMatch {
return r_ch
}
return ""
}
In the file romanization.go, we use it for combind the logic when we just implemented above and return it result to the user.
package romanization
func Romanize(word string) string {
characters := []rune(word)
result := ""
rw := ""
isThreeCh := false
var vCh rune
var lastCh rune
for index, ch := range characters {
// is Suffixable vowel? //ំ, ា, /េ, /ុ ,ោ, /ិ
if IsSuffixableVowel(ch) && vCh == 0 {
vCh = ch
continue
} else if vCh != 0 && (vCh+ch) == 12156 && isThreeCh == false {
// if ាំ special case
isThreeCh = true
continue
}
rw = Roman(ch)
if isThreeCh {
// if ch = ង
if ch == 6020 {
rw = SpecialVowel(12156, ch, IsKorGroup(lastCh))
} else {
rw = SpecialVowel(6070, 6086, IsKorGroup(lastCh)) + rw
}
isThreeCh = false
vCh = 0
} else if vCh != 0 {
rw = SpecialVowel(vCh, ch, IsKorGroup(lastCh))
vCh = 0
} else if IsVowel(ch) {
rw = VowelWord(ch, IsKorGroup(lastCh))
} else if IsConsonant(lastCh) && IsConsonant(ch) {
// ជ + ន = ជ + អr + ន
// ជ + ណ = ជ + អ + ណ
if index == 1 && lastCh == 6050 { //6050 = អ
// if អ at the begin not cound
rw = rw
} else if IsKorGroup(lastCh) {
rw = "o" + rw
} else {
rw = "a" + rw
}
}
lastCh = ch
result = result + rw
}
return result
}
Testing
Now, it done, let run and test it.
./watch.sh
Resources
- Source code
- https://unstats.un.org/unsd/geoinfo/UNGEGN/docs/8th-uncsgn-docs/inf/8th_UNCSGN_econf.94_INF.30_corr1.pdf
- https://en.wiktionary.org/wiki/Wiktionary:Khmer_romanization
- https://www.loc.gov/catdir/cpso/romanization/khmer.pdf
- http://www.eki.ee/wgrs/rom1_km.pdf
Conclussion
Khmer is one of the most complex linguistic system in the world. Khmer Romanization is just a small tip of the iceberg of this complex langauge system. However, when we have an expert at our side, we could implement the algorithm to solve this system with ease. Therefore, I am every thankfull to Mr. Chem Vatho that taking his take for teaching, guiding and supporting me for this chalenge.
All rights reserved
