diff --git a/fluid/PaddleNLP/machine_reading_comprehension/dataset.py b/fluid/PaddleNLP/machine_reading_comprehension/dataset.py
index 9ed3b483cd43ed24b4a60ba6d135c19885989793..c732ce041c5e82ea5e1471ba422f5b056a7cba8f 100644
--- a/fluid/PaddleNLP/machine_reading_comprehension/dataset.py
+++ b/fluid/PaddleNLP/machine_reading_comprehension/dataset.py
@@ -23,6 +23,7 @@ import json
 import logging
 import numpy as np
 from collections import Counter
+import io
 
 
 class BRCDataset(object):
@@ -67,7 +68,7 @@ class BRCDataset(object):
         Args:
             data_path: the data file to load
         """
-        with open(data_path, 'r', encoding='utf-8') as fin:
+        with io.open(data_path, 'r', encoding='utf-8') as fin:
             data_set = []
             for lidx, line in enumerate(fin):
                 sample = json.loads(line.strip())