diff --git a/data/ofa_dataset.py b/data/ofa_dataset.py
index fa30b24c..e3fd31a2 100644
--- a/data/ofa_dataset.py
+++ b/data/ofa_dataset.py
@@ -7,6 +7,11 @@
 import re
 import torch.utils.data
 from fairseq.data import FairseqDataset
+import string       
+
+CHINESE_PUNCTUATION = '＂＃＄％＆＇（）＊＋，－／：；＜＝＞＠［＼］＾＿｀｛｜｝～｟｠｢｣､\u3000、〃〈〉《》「」『』【】〔〕〖〗〘〙〚〛〜〝〞〟〰〾〿–—‘’‛“”„‟…‧﹏﹑﹔·！？｡。'      
+ENGLISH_PUNCTUATION = string.punctuation
+
 
 logger = logging.getLogger(__name__)
 
@@ -77,3 +82,14 @@ def pre_caption(self, caption, max_words=None):
             caption = ' '.join(caption_words[:max_words])
 
         return caption
+
+    def pre_chinese(self, text, max_words):     
+        text = text.lower().replace(CHINESE_PUNCTUATION, " ").replace(ENGLISH_PUNCTUATION, " ")     
+        text = re.sub(      
+            r"\s{2,}",      
+            ' ',        
+            text,       
+        )       
+        text = text.rstrip('\n')        
+        text = text.strip(' ')[:max_words]      
+        return text
diff --git a/data/s2t_data/unify_dataset.py b/data/s2t_data/unify_dataset.py
index 986c1e89..08549b55 100644
--- a/data/s2t_data/unify_dataset.py
+++ b/data/s2t_data/unify_dataset.py
@@ -218,7 +218,6 @@ def __init__(
         self.text2phone_tokenizer = None
         if text2phone_path is not None:
             self.blank_id = self.phone_dict.index("<unk>")
-            self.phone_mask_idx = self.phone_dict.index("#1")
             self.text2phone_tokenizer = Text2Phone(text2phone_path)
 
     def set_epoch(self, epoch, **unused):
diff --git a/tasks/speech_tasks/unify_task.py b/tasks/speech_tasks/unify_task.py
index 1968e515..33b5fa73 100644
--- a/tasks/speech_tasks/unify_task.py
+++ b/tasks/speech_tasks/unify_task.py
@@ -152,7 +152,7 @@ def setup_task(cls, cfg: DictConfig, **kwargs):
         phone_dict = cls.load_dictionary(cfg.phone_dict_path)
         if cfg.text2phone_path is None:
             phone_dict.add_symbol("<blank>")
-        phone_dict.add_symbol("<mask>")
+            phone_dict.add_symbol("<mask>")
 
         # src_dict.add_symbol("<phone_blank>")
         # tgt_dict.add_symbol("<phone_blank>")
@@ -358,6 +358,3 @@ def phone_dictionary(self):
         """Return the phone :class:`~fairseq.data.Dictionary`."""
         return self.phone_dict
 
-    @property
-    def train_stage(self):
-        return self.train_stage
diff --git a/utils/phone/zh/phone_dict.txt b/utils/phone/zh/phone_dict.txt
index 6c775c39..c03774f4 100644
--- a/utils/phone/zh/phone_dict.txt
+++ b/utils/phone/zh/phone_dict.txt
@@ -134,3 +134,4 @@ zh 120
 #14 134
 #15 135
 #16 136
+#17 137