diff --git a/funasr/datasets/large_datasets/utils/tokenize.py b/funasr/datasets/large_datasets/utils/tokenize.py
index 61a1ad001..6ecfb2b63 100644
--- a/funasr/datasets/large_datasets/utils/tokenize.py
+++ b/funasr/datasets/large_datasets/utils/tokenize.py
@@ -59,7 +59,7 @@ def tokenize(data,
             pre_index = None
             for hw in hw_config['pre_hwlist']:
                 hw = " ".join(seg_tokenize(hw, seg_dict))
-                _find = " ".join(text.find(hw))
+                _find = " ".join(text).find(hw)
                 if _find != -1:
                     _find = text[:_find].count(" ")  # bpe sometimes
                     pre_index = [_find, _find + max(hw.count(" "), 1)]