From fa7b072066893966a5309d0863a5aae1eee54dac Mon Sep 17 00:00:00 2001
From: Akito-UzukiP <12011903@mail.sustech.edu.cn>
Date: Sat, 10 Aug 2024 20:09:26 +0800
Subject: [PATCH] add modifiable japanese dict

---
 GPT_SoVITS/text/ja_userdic/userdict.csv |  1 +
 GPT_SoVITS/text/japanese.py             | 24 +++++++++++++++++++++++-
 2 files changed, 24 insertions(+), 1 deletion(-)
 create mode 100644 GPT_SoVITS/text/ja_userdic/userdict.csv

diff --git a/GPT_SoVITS/text/ja_userdic/userdict.csv b/GPT_SoVITS/text/ja_userdic/userdict.csv
new file mode 100644
index 00000000..b23e0d63
--- /dev/null
+++ b/GPT_SoVITS/text/ja_userdic/userdict.csv
@@ -0,0 +1 @@
+主殿,*,*,-32767,名詞,固有名詞,一般,*,*,*,アルジドノ,アルジドノ,アルジドノ,3/5,*
\ No newline at end of file
diff --git a/GPT_SoVITS/text/japanese.py b/GPT_SoVITS/text/japanese.py
index 4c10720e..e9fe7c1a 100644
--- a/GPT_SoVITS/text/japanese.py
+++ b/GPT_SoVITS/text/japanese.py
@@ -1,8 +1,30 @@
 # modified from https://github.com/CjangCjengh/vits/blob/main/text/japanese.py
 import re
-import sys
 
 import pyopenjtalk
+import os
+import hashlib
+current_file_path = os.path.dirname(__file__)
+def get_hash(fp: str) -> str:
+    hash_md5 = hashlib.md5()
+    with open(fp, "rb") as f:
+        for chunk in iter(lambda: f.read(4096), b""):
+            hash_md5.update(chunk)
+    return hash_md5.hexdigest()
+
+USERDIC_CSV_PATH = os.path.join(current_file_path, "ja_userdic", "userdict.csv")
+USERDIC_BIN_PATH = os.path.join(current_file_path, "ja_userdic", "user.dict")
+USERDIC_HASH_PATH = os.path.join(current_file_path, "ja_userdic", "userdict.md5")
+# 如果没有用户词典，就生成一个；如果有，就检查md5，如果不一样，就重新生成
+if os.path.exists(USERDIC_CSV_PATH):
+    if not os.path.exists(USERDIC_BIN_PATH) or get_hash(USERDIC_CSV_PATH) != open(USERDIC_HASH_PATH, "r",encoding='utf-8').read():
+        pyopenjtalk.mecab_dict_index(USERDIC_CSV_PATH, USERDIC_BIN_PATH)
+        with open(USERDIC_HASH_PATH, "w", encoding='utf-8') as f:
+            f.write(get_hash(USERDIC_CSV_PATH))
+
+if os.path.exists(USERDIC_BIN_PATH):
+    pyopenjtalk.update_global_jtalk_with_user_dict(USERDIC_BIN_PATH)    
+
 
 from text.symbols import punctuation
 # Regular expression matching Japanese without punctuation marks: