diff --git a/requirements/base.in b/requirements/base.in
index cc2b27d8ad..09c1dc196d 100644
--- a/requirements/base.in
+++ b/requirements/base.in
@@ -22,3 +22,4 @@ tqdm
 psutil
 python-oxmsg
 html5lib
+chardet
diff --git a/unstructured/partition/common/common.py b/unstructured/partition/common/common.py
index 267630a87b..9c4d492753 100644
--- a/unstructured/partition/common/common.py
+++ b/unstructured/partition/common/common.py
@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+import chardet
 import numbers
 import subprocess
 from io import BufferedReader, BytesIO, TextIOWrapper
@@ -296,7 +297,9 @@ def convert_office_doc(
         wait_time = 0
         sleep_time = 0.1
         output = subprocess.run(command, capture_output=True)
-        message = output.stdout.decode().strip()
+        detected_encoding = chardet.detect(output.stdout)
+        encoding = detected_encoding['encoding'] or 'utf-8'  # Default to utf-8 if detection fails
+        message = output.stdout.decode(encoding).strip()
         # we can't rely on returncode unfortunately because on macOS it would return 0 even when the
         # command failed to run; instead we have to rely on the stdout being empty as a sign of the
         # process failed