phoonnx 0.0.0__py3-none-any.whl

phoonnx/thirdparty/ko_tables/ipa.csv

@@ -0,0 +1,22 @@
+C,CSymbol,V,VSymbol
+ㅂ,p,ㅏ,ä
+ㄷ,t,ㅔ,e
+ㅌ,tʰ,ㅐ,ɛ
+ㅈ,tɕ,ㅣ,i
+ㅉ,tɕ*,ㅗ,o
+ㅊ,tɕʰ,ㅚ,wɛ
+ㄱ,k,ㅜ,u
+ㅎ,h,ㅓ,ʌ̹
+ㄲ,k*,ㅡ,ɯ
+ㅋ,kʰ,ㅢ,ɰi
+ㄹ,l,ㅛ,jo
+ㅁ,m,ㅠ,ju
+ㄴ,n,ㅑ,ja
+ㅇ,ŋ,ㅕ,jʌ̹
+ㄸ,t*,ㅖ,je
+ㅃ,p*,ㅒ,jɛ
+ㅍ,pʰ,ㅘ,wa
+ㅅ,s,ㅝ,wʌ̹
+ㅆ,s*,ㅟ,wi
+#,#,ㅙ,wɛ
+$,$,ㅞ,we

phoonnx/thirdparty/ko_tables/neutralization.csv

@@ -0,0 +1,11 @@
+_from,_to
+ㄲ,ㄱ
+ㅋ,ㄱ
+ㅅ,ㄷ
+ㅆ,ㄷ
+ㅈ,ㄷ
+ㅊ,ㄷ
+ㅌ,ㄷ
+ㅎ,ㄷ
+ㅂ,ㅂ
+ㅍ,ㅂ

phoonnx/thirdparty/ko_tables/tensification.csv

@@ -0,0 +1,56 @@
+_from,_to
+ㄱㄱ,ㄱㄲ
+ㄲㄱ,ㄲㄲ
+ㅋㄱ,ㅋㄲ
+ㄱㄷ,ㄱㄸ
+ㄲㄷ,ㄲㄸ
+ㅋㄷ,ㅋㄸ
+ㄱㅂ,ㄱㅃ
+ㄲㅂ,ㄲㅃ
+ㅋㅂ,ㅋㅃ
+ㄱㅅ,ㄱㅆ
+ㄲㅅ,ㄲㅆ
+ㅋㅅ,ㅋㅆ
+ㄱㅈ,ㄱㅉ
+ㄲㅈ,ㄲㅉ
+ㅋㅈ,ㅋㅉ
+ㄷㄱ,ㄷㄲ
+ㅅㄱ,ㅅㄲ
+ㅆㄱ,ㅆㄲ
+ㅈㄱ,ㅈㄲ
+ㅊㄱ,ㅊㄲ
+ㅌㄱ,ㅌㄲ
+ㄷㄷ,ㄷㄸ
+ㅅㄷ,ㅅㄸ
+ㅆㄷ,ㅆㄸ
+ㅈㄷ,ㅈㄸ
+ㅊㄷ,ㅊㄸ
+ㅌㄷ,ㅌㄸ
+ㄷㅂ,ㄷㅃ
+ㅅㅂ,ㅅㅃ
+ㅆㅂ,ㅆㅃ
+ㅈㅂ,ㅈㅃ
+ㅊㅂ,ㅊㅃ
+ㅌㅂ,ㅌㅃ
+ㄷㅅ,ㄷㅆ
+ㅅㅅ,ㅅㅆ
+ㅆㅅ,ㅆㅆ
+ㅈㅅ,ㅈㅆ
+ㅊㅅ,ㅊㅆ
+ㅌㅅ,ㅌㅆ
+ㄷㅈ,ㄷㅉ
+ㅅㅈ,ㅅㅉ
+ㅆㅈ,ㅆㅉ
+ㅈㅈ,ㅈㅉ
+ㅊㅈ,ㅊㅉ
+ㅌㅈ,ㅌㅉ
+ㅂㄱ,ㅂㄲ
+ㅍㄱ,ㅍㄲ
+ㅂㄷ,ㅂㄸ
+ㅍㄷ,ㅍㄸ
+ㅂㅂ,ㅂㅃ
+ㅍㅂ,ㅍㅃ
+ㅂㅅ,ㅂㅆ
+ㅍㅅ,ㅍㅆ
+ㅂㅈ,ㅂㅉ
+ㅍㅈ,ㅍㅉ

phoonnx/thirdparty/ko_tables/yale.csv

@@ -0,0 +1,22 @@
+C,CSymbol,V,VSymbol
+ㅂ,p,ㅏ,a
+ㄷ,t,ㅔ,ey
+ㅌ,th,ㅐ,ay
+ㅈ,c,ㅣ,i
+ㅉ,cc,ㅗ,o
+ㅊ,ch,ㅚ,oy
+ㄱ,k,ㅜ,wu
+ㅎ,h,ㅓ,e
+ㄲ,kk,ㅡ,u
+ㅋ,kh,ㅢ,uy
+ㄹ,l,ㅛ,yo
+ㅁ,m,ㅠ,yu
+ㄴ,n,ㅑ,ya
+ㅇ,ng,ㅕ,ye
+ㄸ,tt,ㅖ,yey
+ㅃ,pp,ㅒ,yay
+ㅍ,ph,ㅘ,wa
+ㅅ,s,ㅝ,we
+ㅆ,ss,ㅟ,wi
+#,#,ㅙ,way
+$,$,ㅞ,wey

phoonnx/thirdparty/kog2p/__init__.py

@@ -0,0 +1,385 @@
+# taken from https://github.com/scarletcho/KoG2P
+'''
+g2p.py
+~~~~~~~~~~
+
+This script converts Korean graphemes to romanized phones and then to pronunciation.
+
+    (1) graph2phone: convert Korean graphemes to romanized phones
+    (2) phone2prono: convert romanized phones to pronunciation
+    (3) graph2phone: convert Korean graphemes to pronunciation
+
+Usage:  $ python g2p.py '스물 여덟째 사람'
+        (NB. Please check 'rulebook_path' before usage.)
+
+Yejin Cho (ycho@utexas.edu)
+Jaegu Kang (jaekoo.jk@gmail.com)
+Hyungwon Yang (hyung8758@gmail.com)
+Yeonjung Hong (yvonne.yj.hong@gmail.com)
+
+Created: 2016-08-11
+Last updated: 2019-01-31 Yejin Cho
+
+* Key updates made:
+    - Executable in both Python 2 and 3.
+    - G2P Performance test available ($ python g2p.py test)
+    - G2P verbosity control available
+
+'''
+
+import datetime as dt
+import os.path
+import re
+import math
+import sys
+import optparse
+
+# Option
+parser = optparse.OptionParser()
+parser.add_option("-v", action="store_true", dest="verbose", default="False",
+                  help="This option prints the detail information of g2p process.")
+
+(options, args) = parser.parse_args()
+verbose = options.verbose
+
+# Check Python version
+ver_info = sys.version_info
+
+if ver_info[0] == 2:
+    reload(sys)
+    sys.setdefaultencoding('utf-8')
+
+
+def readfileUTF8(fname):
+    f = open(fname, 'r')
+    corpus = []
+
+    while True:
+        line = f.readline()
+        line = line.encode("utf-8")
+        line = re.sub(u'\n', u'', line)
+        if line != u'':
+            corpus.append(line)
+        if not line: break
+
+    f.close()
+    return corpus
+
+
+def writefile(body, fname):
+    out = open(fname, 'w')
+    for line in body:
+        out.write('{}\n'.format(line))
+    out.close()
+
+
+def readRules(pver, rule_book):
+    if pver == 2:
+        f = open(rule_book, 'r')
+    elif pver == 3:
+        f = open(rule_book, 'r', encoding="utf-8")
+
+    rule_in = []
+    rule_out = []
+
+    while True:
+        line = f.readline()
+        if pver == 2:
+            line = unicode(line.encode("utf-8"))
+            line = re.sub(u'\n', u'', line)
+        elif pver == 3:
+            line = re.sub('\n', '', line)
+
+        if line != u'':
+            if line[0] != u'#':
+                IOlist = line.split('\t')
+                rule_in.append(IOlist[0])
+                if IOlist[1]:
+                    rule_out.append(IOlist[1])
+                else:  # If output is empty (i.e. deletion rule)
+                    rule_out.append(u'')
+        if not line: break
+    f.close()
+
+    return rule_in, rule_out
+
+
+def isHangul(charint):
+    hangul_init = 44032
+    hangul_fin = 55203
+    return charint >= hangul_init and charint <= hangul_fin
+
+
+def checkCharType(var_list):
+    #  1: whitespace
+    #  0: hangul
+    # -1: non-hangul
+    checked = []
+    for i in range(len(var_list)):
+        if var_list[i] == 32:  # whitespace
+            checked.append(1)
+        elif isHangul(var_list[i]):  # Hangul character
+            checked.append(0)
+        else:  # Non-hangul character
+            checked.append(-1)
+    return checked
+
+
+def graph2phone(graphs):
+    # Encode graphemes as utf8
+    try:
+        graphs = graphs.decode('utf8')
+    except AttributeError:
+        pass
+
+    integers = []
+    for i in range(len(graphs)):
+        integers.append(ord(graphs[i]))
+
+    # Romanization (according to Korean Spontaneous Speech corpus; 성인자유발화코퍼스)
+    phones = ''
+    ONS = ['k0', 'kk', 'nn', 't0', 'tt', 'rr', 'mm', 'p0', 'pp',
+           's0', 'ss', 'oh', 'c0', 'cc', 'ch', 'kh', 'th', 'ph', 'h0']
+    NUC = ['aa', 'qq', 'ya', 'yq', 'vv', 'ee', 'yv', 'ye', 'oo', 'wa',
+           'wq', 'wo', 'yo', 'uu', 'wv', 'we', 'wi', 'yu', 'xx', 'xi', 'ii']
+    COD = ['', 'kf', 'kk', 'ks', 'nf', 'nc', 'nh', 'tf',
+           'll', 'lk', 'lm', 'lb', 'ls', 'lt', 'lp', 'lh',
+           'mf', 'pf', 'ps', 's0', 'ss', 'oh', 'c0', 'ch',
+           'kh', 'th', 'ph', 'h0']
+
+    # Pronunciation
+    idx = checkCharType(integers)
+    iElement = 0
+    while iElement < len(integers):
+        if idx[iElement] == 0:  # not space characters
+            base = 44032
+            df = int(integers[iElement]) - base
+            iONS = int(math.floor(df / 588)) + 1
+            iNUC = int(math.floor((df % 588) / 28)) + 1
+            iCOD = int((df % 588) % 28) + 1
+
+            s1 = '-' + ONS[iONS - 1]  # onset
+            s2 = NUC[iNUC - 1]  # nucleus
+
+            if COD[iCOD - 1]:  # coda
+                s3 = COD[iCOD - 1]
+            else:
+                s3 = ''
+            tmp = s1 + s2 + s3
+            phones = phones + tmp
+
+        elif idx[iElement] == 1:  # space character
+            tmp = '#'
+            phones = phones + tmp
+
+        phones = re.sub('-(oh)', '-', phones)
+        iElement += 1
+        tmp = ''
+
+    # 초성 이응 삭제
+    phones = re.sub('^oh', '', phones)
+    phones = re.sub('-(oh)', '', phones)
+
+    # 받침 이응 'ng'으로 처리 (Velar nasal in coda position)
+    phones = re.sub('oh-', 'ng-', phones)
+    phones = re.sub('oh([# ]|$)', 'ng', phones)
+
+    # Remove all characters except Hangul and syllable delimiter (hyphen; '-')
+    phones = re.sub('(\W+)\-', '\\1', phones)
+    phones = re.sub('\W+$', '', phones)
+    phones = re.sub('^\-', '', phones)
+    return phones
+
+
+def phone2prono(phones, rule_in, rule_out):
+    # Apply g2p rules
+    for pattern, replacement in zip(rule_in, rule_out):
+        # print pattern
+        phones = re.sub(pattern, replacement, phones)
+        prono = phones
+    return prono
+
+
+def addPhoneBoundary(phones):
+    # Add a comma (,) after every second alphabets to mark phone boundaries
+    ipos = 0
+    newphones = ''
+    while ipos + 2 <= len(phones):
+        if phones[ipos] == u'-':
+            newphones = newphones + phones[ipos]
+            ipos += 1
+        elif phones[ipos] == u' ':
+            ipos += 1
+        elif phones[ipos] == u'#':
+            newphones = newphones + phones[ipos]
+            ipos += 1
+
+        newphones = newphones + phones[ipos] + phones[ipos + 1] + u','
+        ipos += 2
+
+    return newphones
+
+
+def addSpace(phones):
+    ipos = 0
+    newphones = ''
+    while ipos < len(phones):
+        if ipos == 0:
+            newphones = newphones + phones[ipos] + phones[ipos + 1]
+        else:
+            newphones = newphones + ' ' + phones[ipos] + phones[ipos + 1]
+        ipos += 2
+
+    return newphones
+
+
+def graph2prono(graphs, rule_in, rule_out):
+    romanized = graph2phone(graphs)
+    romanized_bd = addPhoneBoundary(romanized)
+    prono = phone2prono(romanized_bd, rule_in, rule_out)
+
+    prono = re.sub(u',', u' ', prono)
+    prono = re.sub(u' $', u'', prono)
+    prono = re.sub(u'#', u'-', prono)
+    prono = re.sub(u'-+', u'-', prono)
+
+    prono_prev = prono
+    identical = False
+    loop_cnt = 1
+
+    if verbose == True:
+        print('=> Romanized: ' + romanized)
+        print('=> Romanized with boundaries: ' + romanized_bd)
+        print('=> Initial output: ' + prono)
+
+    while not identical:
+        prono_new = phone2prono(re.sub(u' ', u',', prono_prev + u','), rule_in, rule_out)
+        prono_new = re.sub(u',', u' ', prono_new)
+        prono_new = re.sub(u' $', u'', prono_new)
+
+        if re.sub(u'-', u'', prono_prev) == re.sub(u'-', u'', prono_new):
+            identical = True
+            prono_new = re.sub(u'-', u'', prono_new)
+            if verbose == True:
+                print('\n=> Exhaustive rule application completed!')
+                print('=> Total loop count: ' + str(loop_cnt))
+                print('=> Output: ' + prono_new)
+        else:
+            if verbose == True:
+                print('\n=> Rule applied for more than once')
+                print('cmp1: ' + re.sub(u'-', u'', prono_prev))
+                print('cmp2: ' + re.sub(u'-', u'', prono_new))
+            loop_cnt += 1
+            prono_prev = prono_new
+
+    return prono_new
+
+
+def testG2P(rulebook, testset):
+    [testin, testout] = readRules(ver_info[0], testset)
+    cnt = 0
+    body = []
+    for idx in range(0, len(testin)):
+        print('Test item #: ' + str(idx + 1) + '/' + str(len(testin)))
+        item_in = testin[idx]
+        item_out = testout[idx]
+        ans = graph2phone(item_out)
+        ans = re.sub(u'-', u'', ans)
+        ans = addSpace(ans)
+
+        [rule_in, rule_out] = readRules(ver_info[0], rulebook)
+        pred = graph2prono(item_in, rule_in, rule_out)
+
+        if pred != ans:
+            print('G2P ERROR:  [result] ' + pred + '\t\t\t[ans] ' + item_in + ' [' + item_out + '] ' + ans)
+            cnt += 1
+        else:
+            body.append('[result] ' + pred + '\t\t\t[ans] ' + item_in + ' [' + item_out + '] ' + ans)
+
+    print('Total error item #: ' + str(cnt))
+    writefile(body, 'good.txt')
+
+
+kog2p_to_hangul = {
+    # Onset consonants
+    'p0': 'ㅂ',
+    'ph': 'ㅍ',
+    'pp': 'ㅃ',
+    't0': 'ㄷ',
+    'th': 'ㅌ',
+    'tt': 'ㄸ',
+    'k0': 'ㄱ',
+    'kh': 'ㅋ',
+    'kk': 'ㄲ',
+    's0': 'ㅅ',
+    'ss': 'ㅆ',
+    'h0': 'ㅎ',
+    'c0': 'ㅈ',
+    'ch': 'ㅊ',
+    'cc': 'ㅉ',
+    'mm': 'ㅁ',
+    'nn': 'ㄴ',
+    'rr': 'ㄹ',
+
+    # Coda consonants
+    'pf': 'ㅂ',
+    'tf': 'ㄷ',
+    'kf': 'ㄱ',
+    'mf': 'ㅁ',
+    'nf': 'ㄴ',
+    'ng': 'ㅇ',
+    'll': 'ㄹ',
+    'ks': 'ㄱㅅ',
+    'nc': 'ㄴㅈ',
+    'nh': 'ㄴㅎ',
+    'lk': 'ㄹㄱ',
+    'lm': 'ㄹㅁ',
+    'lb': 'ㄹㅂ',
+    'ls': 'ㄹㅅ',
+    'lt': 'ㄹㅌ',
+    'lp': 'ㄹㅍ',
+    'lh': 'ㄹㅎ',
+    'ps': 'ㅂㅅ',
+
+    # Monophthongs (vowels)
+    'ii': 'ㅣ',
+    'ee': 'ㅔ',
+    'qq': 'ㅐ',
+    'aa': 'ㅏ',
+    'xx': 'ㅡ',
+    'vv': 'ㅓ',
+    'uu': 'ㅜ',
+    'oo': 'ㅗ',
+
+    # Diphthongs (vowels)
+    'ye': 'ㅖ',
+    'yq': 'ㅒ',
+    'ya': 'ㅑ',
+    'yv': 'ㅕ',
+    'yu': 'ㅠ',
+    'yo': 'ㅛ',
+    'wi': 'ㅟ',
+    'wo': 'ㅚ',
+    'wq': 'ㅙ',
+    'we': 'ㅞ',
+    'wa': 'ㅘ',
+    'wv': 'ㅝ',
+    'xi': 'ㅢ',
+}
+
+
+def runKoG2P(graph, rulebook=None):
+    if not rulebook:
+        rulebook = f"{os.path.dirname(__file__)}/rulebook.txt"
+    [rule_in, rule_out] = readRules(ver_info[0], rulebook)
+    words = graph.split()
+    phonemized = []
+    for w in words:
+        phones = graph2prono(w, rule_in, rule_out).split()
+        hangul = "".join([kog2p_to_hangul[p] for p in phones])
+        phonemized.append(hangul)
+    return " ".join(phonemized)
+
+
+
+