sigilyph 0.5.2__py3-none-any.whl

sigilyph/core/sigilyph_class.py

@@ -0,0 +1,246 @@
+'''
+FilePath: /python-Sigilyph/sigilyph/core/sigilyph_class.py
+Descripttion: 
+Author: Yixiang Chen
+version: 
+Date: 2025-08-12 14:42:50
+LastEditors: Yixiang Chen
+LastEditTime: 2026-01-16 18:45:50
+'''
+
+import langid
+import re
+import json
+
+import jieba
+import jieba.posseg
+
+from sigilyph.core.g2p_func import g2p_en, g2p_cn
+from sigilyph.text_norm.norm_func import preprocess_first, post_process, text_norm_en, text_norm_cn
+from sigilyph.core.symbols import punctuation 
+from sigilyph.core.predict import before_replace_dict, special_word_dict, special_phrase
+
+norm_func_dict = {
+    'en': text_norm_en,
+    'zh': text_norm_cn
+}
+
+g2p_func_dict = {
+    'en': g2p_en,
+    'zh': g2p_cn
+}
+
+class Sigilyph:
+    def __init__(self, before_dict_path=None, special_dict_path=None):
+        self.sil1symbol='-'
+        self.punctuation = punctuation
+
+        self.before_replace_dict = before_replace_dict 
+        if before_dict_path:
+            with open(before_dict_path, 'r', encoding="utf-8") as obdp:
+                extra_before_dict = json.load(obdp)
+            self.before_replace_dict.update(extra_before_dict)
+
+        self.special_word_dict = special_word_dict
+        if special_dict_path:
+            with open(special_dict_path, 'r', encoding="utf-8") as obdp:
+                extra_special_dict = json.load(obdp)
+            self.special_word_dict.update(extra_special_dict)
+        
+        self.special_phrase = special_phrase 
+    
+    def forward(self, text, lang, spflag=False, norm_use_lang='zh'):
+        phones = self.text_process(text, lang, spflag, norm_use_lang)
+        phones = self.replace_sil2label(phones)
+        return phones
+
+    def text_process(self, text, lang, spflag=False, norm_use_lang='zh'):
+        text = preprocess_first(text, self.before_replace_dict, special_word_dict, norm_use_lang=norm_use_lang)
+        text = self.text_norm(use_text, norm_use_lang) 
+        text = post_process(text, special_word_dict)
+
+        multi_lang_text_list = self.text_split_lang(text, lang) 
+
+        all_phone = []
+        for text_split_dict in multi_lang_text_list:
+            use_lang = text_split_dict['lang']
+            use_text = text_split_dict['text_split']
+            if use_lang == 'phone':
+                phonelist = use_text.split()
+                all_phone.extend(phonelist) 
+            else:
+                if use_lang not in norm_func_dict.keys():
+                    use_lang = 'zh'
+                #use_text = self.text_norm(use_text, use_lang)
+                use_text = use_text
+                phone_list = self.g2p(use_text, use_lang)
+                #all_phone.append('sil')
+                all_phone.append('sil_lang')
+                all_phone.append('<sp>')
+                all_phone.extend(phone_list)
+        #all_phone = postprocess(all_phone)
+        all_phone = self.postprocess_tts(all_phone)
+        if not spflag:
+            while '<sp>' in all_phone:
+                all_phone.remove('<sp>')
+        return all_phone
+
+    def text_process_old(self, text, lang, spflag=False, norm_use_lang='zh'):
+        text = preprocess_first(text, self.before_replace_dict, special_word_dict, norm_use_lang=norm_use_lang)
+
+        multi_lang_text_list = self.text_split_lang(text, lang) 
+
+        all_phone = []
+        for text_split_dict in multi_lang_text_list:
+            use_lang = text_split_dict['lang']
+            use_text = text_split_dict['text_split']
+            if use_lang == 'phone':
+                phonelist = use_text.split()
+                all_phone.extend(phonelist) 
+            else:
+                if use_lang not in norm_func_dict.keys():
+                    use_lang = 'zh'
+                use_text = self.text_norm(use_text, use_lang)
+                phone_list = self.g2p(use_text, use_lang)
+                #all_phone.append('sil')
+                all_phone.append('sil_lang')
+                all_phone.append('<sp>')
+                all_phone.extend(phone_list)
+        #all_phone = postprocess(all_phone)
+        all_phone = self.postprocess_tts(all_phone)
+        if not spflag:
+            while '<sp>' in all_phone:
+                all_phone.remove('<sp>')
+        return all_phone
+
+    ###############  split text in line with lang ##############
+    def text_split_lang(self, text, lang):
+        if lang == 'ZH' or lang == 'zh':
+            multi_lang_text_list = [{'lang':'zh', 'text_split': text}]
+        elif lang == 'en':
+            multi_lang_text_list = [{'lang':'en', 'text_split': text}]
+        else:
+            pretext_split =  re.split("(\[.*?\])", text, re.I|re.M)
+            multi_lang_text_list = []
+            pretext_split = list(filter(None, pretext_split))
+            for utext in pretext_split:
+                if utext[0] != '[':
+                    pattern = r'([a-zA-Z ,.\!\?]+|[\u4e00-\u9fa5 ，。,.\t \"\！\？\“\”\、]+)'
+                    text_split = re.findall(pattern, utext)
+                    #print(text_split)
+                    for idx in range(len(text_split)):
+                        tmpts = text_split[idx]
+                        tmp_lang = langid.classify(tmpts)[0]
+                        if len(tmpts)>20:
+                            if not self.has_punc(tmpts[:-1]):
+                                tmpts = self.add_pause(tmpts, 'p')
+                            if not self.has_punc(tmpts[:-1]):
+                                tmpts = self.add_pause(tmpts, 'v')   
+                        if tmpts in self.special_phrase:
+                            tmpts = tmpts+self.sil1symbol
+                        if tmp_lang in ['zh', 'jp', 'ja']:
+                            tmp_lang = 'zh'
+                            tmpts = tmpts.replace(' ', self.sil1symbol)
+                        else:
+                            tmp_lang = 'en' 
+                        if not tmpts.isspace():
+                            multi_lang_text_list.append({'lang':tmp_lang, 'text_split': tmpts})
+                else:
+                    phones = utext[1:-1]
+                    multi_lang_text_list.append({'lang':'phone', 'text_split': phones})
+        return multi_lang_text_list
+    
+    ##########  add parse ###############
+    def has_punc(self, text):
+        for char in text:
+            if char in [',', '.', '!', '?', '，','。','？','！', self.sil1symbol]:
+                return True
+        return False
+    
+    def add_pause(self, text, tf='v'):
+        segment = jieba.posseg.cut(text.strip())
+        wlist = []
+        flist = []
+        for x in segment:
+            wlist.append(x.word)
+            flist.append(x.flag)
+        idx = self.search_ele_mid(flist, tf)
+        if idx != len(flist)-1:
+            wlist.insert(idx, self.sil1symbol)
+        outtext = ''.join(wlist)
+        return outtext
+    
+    def search_ele_mid(self, flaglist, tf = 'v'):
+        nowidx = -1
+        halflen = (len(flaglist))//2
+        for gap in range(len(flaglist)-halflen):
+            nowidx = halflen - gap
+            if flaglist[nowidx]==tf:
+                return nowidx
+            nowidx = halflen + gap
+            if flaglist[nowidx]==tf:
+                return nowidx
+        return nowidx
+
+    ######## text norm #########
+    def text_norm(self, text, lang):
+        outtext = norm_func_dict[lang](text)
+        return outtext
+
+    ############ g2p ################
+    def g2p(self, text, lang):
+        phoneme_list = g2p_func_dict[lang](text)
+        return phoneme_list
+    
+    ############# post process #############
+    def postprocess_tts(self, phonelist):
+        #outlist = ['sil', '<sp>']
+        outlist = []
+        #print(phonelist)
+        for idx in range(len(phonelist)):
+            pm = phonelist[idx]
+            if pm not in self.punctuation:
+                outlist.append(pm)
+            elif pm == self.sil1symbol:
+                outlist.append('sil_1')
+            else:
+                #outlist.append('sil')
+                outlist.append('sil_punc')
+                #outlist.append('<sp>')
+        #if outlist[-1] == 'sil': 
+        #    outlist.append('<sp>')
+        #elif outlist[-2] != 'sil':
+        #    outlist.append('sil')
+        #    outlist.append('<sp>')
+        if phonelist[-2] not in self.punctuation and outlist[-1].split('_')[0] != 'sil':
+            #outlist.append('sil')
+            outlist.append('sil_end')
+            outlist.append('<sp>')
+        return outlist
+
+    ########## replace silence token ###############
+    def replace_sil2label(self, phones):
+        #phones = ['sil_1' if xx == 'sil_lang' else xx for xx in phones]
+        phones = ['' if xx == 'sil_lang' else xx for xx in phones]
+        phones = ['sil_2' if xx == 'sil_punc' else xx for xx in phones]
+        phones = ['sil_2' if xx == 'sil_end' else xx for xx in phones]
+        phones = ['sil_1' if xx == 'sil' else xx for xx in phones]
+        phones = list(filter(None, phones))
+        #outphones = []
+        outphones = ['sil_1']
+        for ele in phones:
+            if outphones == []:
+                outphones.append(ele)
+            else:
+                if ele.split('_')[0] == 'sil' and outphones[-1].split('_')[0] == 'sil':
+                    outphones[-1] = 'sil_2'
+                    #outphones[-1] = 'sil_1'
+                else:
+                    outphones.append(ele)
+        #if outphones[-1].split('_')[0] == 'sil':
+        #    outphones = outphones[:-1]
+        return outphones
+    
+    
+    
+

sigilyph/core/special_dict.json

@@ -0,0 +1,26 @@
+{
+  "甄嬛传": "[zh en1 h uan2 zh uan4]",
+  "藏海传": "[z ang4 h ai3 zh uan4]",
+  "藏海": "[z ang4 h ai3]",
+  "井柏然": "[j ing3 b o2 r an2]",
+  "Alibaba": "[AA a1 l i3 b a1 b a1]",
+  "TFBOYS": "[T IY1 EH1 F B OY1 Z]",
+  "肖战": "[x iao1 zh an4]",
+  "肖申克": "[x iao1 sh en1 k e4]",
+  "肖像": "[x iao1 x iang4]",
+  "戛纳": "[g a1 n a4]",
+  "爆肚": "[b ao4 d u3 EE]",
+  "花呗": "[h ua1 b ei4]",
+  "肚儿": "[d u3 er5]",
+  "种了": "[zh ong4 l e5]",
+  "仇老五": "[q iu2 l ao3 w u3]",
+  "彩云曲": "[c ai3 y vn2 q v3]",
+  "雷洛传": "[l ei2 l uo4 zh uan4]",
+  "朝雪录": "[zh ao1 x ve3 l u4]",
+  "凡人修仙传": "[f an2 r en2 x iu1 x ian1 zh uan4]",
+  "+86": "[b a1 l iu]",
+  "（正八十六）": "[b a1 l iu]",
+  "八 a.m.": "[sh ang4 w u3 b a1 d ian3]",
+  "F.I.R.": "[EH1 F sil_1 AY2 sil_1 AA0 R sil_1]",
+  "S.C.I": "[sil_2 EH2 S sil_1 S IY1 sil_1 AY1 sil_1]"
+}