diff --git a/egs/gale_arabic/s5d/local/check_tools.sh b/egs/gale_arabic/s5d/local/check_tools.sh
new file mode 100755
index 00000000000..448a6536946
--- /dev/null
+++ b/egs/gale_arabic/s5d/local/check_tools.sh
@@ -0,0 +1,15 @@
+#!/bin/bash
+
+# check whether bs4 and lxml is installed
+if ! python3 -c "import bs4" 2>/dev/null; then
+  echo "$0: BeautifulSoup4 not installed, you can install it by 'pip install beautifulsoup4' if you prefer to use python to process xml file" 
+  exit 1;
+fi
+
+if ! python3 -c "import lxml" 2>/dev/null; then
+  echo "$0: lxml not installed, you can install it by 'pip install lxml' if you prefer to use python to process xml file"
+  exit 1;
+fi
+
+echo "both BeatufileSoup4 and lxml are installed in python"
+exit 0
diff --git a/egs/gale_arabic/s5d/local/prepare_data.sh b/egs/gale_arabic/s5d/local/prepare_data.sh
index 820f938b267..d09ff00acd0 100755
--- a/egs/gale_arabic/s5d/local/prepare_data.sh
+++ b/egs/gale_arabic/s5d/local/prepare_data.sh
@@ -31,6 +31,7 @@ dir10=/export/corpora/LDC/LDC2018S05/
 text10=/export/corpora/LDC/LDC2018T14/        
 
 mgb2_dir=""
+process_xml=""
 mer=80
 
 . ./utils/parse_options.sh
@@ -108,13 +109,6 @@ cd $top_pwd
 # prepare MGB2 data
 if [ ! -z $mgb2_dir ]; then
   echo "preparing MGB2 data"
-  # check xml
-  if [ -z $(which xml) ]; then
-    echo "$0: Could not find tool xml"
-    echo "$0: To use MGB2 you must have xml installed"
-    echo "$0: Download and install it from xmlstar.sourceforge.net"
-    exit 1
-  fi
 
   xmldir=$mgb2_dir/train/xml/bw
   output_dir=$gale_data/mgb2
@@ -126,12 +120,35 @@ if [ ! -z $mgb2_dir ]; then
     mv $output_dir/mgb2 ${output_dir}/.backup
   fi
 
-  ls $mgb2_dir/train/wav/ | while read name; do
-    basename=`basename -s .wav $name`
-    [ ! -e $xmldir/$basename.xml ] && echo "Missing $xmldir/$basename.xml" && exit 1
-    xml sel -t -m '//segments[@annotation_id="transcript_align"]' -m "segment" -n -v  "concat(@who,' ',@starttime,' ',@endtime,' ',@WMER,' ')" -m "element" -v "concat(text(),' ')" $xmldir/$basename.xml | local/add_to_datadir.py $basename $output_dir $mer
-     echo $basename $db_dir/train/wav/$basename.wav >> $output_dir/wav.scp
-  done
+  if [ $process_xml == 'python' ]; then
+    echo "using python to process xml file"
+    # check if bs4 and lxml are installed in python
+    local/check_tools.sh
+    ls $mgb2_dir/train/wav/ | while read name; do
+      basename=`basename -s .wav $name`
+      [ ! -e $xmldir/$basename.xml ] && echo "Missing $xmldir/$basename.xml" && exit 1
+      local/process_xml.py $xmldir/$basename.xml - | local/add_to_datadir.py $basename $train_dir $mer
+      echo $basename $db_dir/train/wav/$basename.wav >> $output_dir/wav.scp
+    done
+  elif [ $process_xml == 'xml' ]; then
+    # check if xml binary exsits
+    if command -v xml >/dev/null 2>/dev/null; then
+      echo "using xml"
+      ls $mgb2_dir/train/wav/ | while read name; do
+        basename=`basename -s .wav $name`
+        [ ! -e $xmldir/$basename.xml ] && echo "Missing $xmldir/$basename.xml" && exit 1
+        xml sel -t -m '//segments[@annotation_id="transcript_align"]' -m "segment" -n -v  "concat(@who,' ',@starttime,' ',@endtime,' ',@WMER,' ')" -m "element" -v "concat(text(),' ')" $xmldir/$basename.xml | local/add_to_datadir.py $basename $output_dir $mer
+        echo $basename $db_dir/train/wav/$basename.wav >> $output_dir/wav.scp
+      done
+    else
+      echo "xml not found, you may use python by '--process-xml python'"
+      exit 1;
+    fi
+  else
+    # invalid option
+    echo "$0: invalid option for --process-xml, choose from 'xml' or 'python'"
+    exit 1;
+  fi
 
   # add mgb2 data to training data (GALE/all and wav.scp)
   mv $gale_data/all $gale_data/all.gale 
diff --git a/egs/gale_arabic/s5d/local/process_xml.py b/egs/gale_arabic/s5d/local/process_xml.py
new file mode 100755
index 00000000000..3c6eed452ac
--- /dev/null
+++ b/egs/gale_arabic/s5d/local/process_xml.py
@@ -0,0 +1,35 @@
+#!/usr/bin/env python3
+
+from bs4 import BeautifulSoup
+import sys
+import argparse
+
+def get_args():
+    parser = argparse.ArgumentParser(description="""This script process xml file.""")
+    parser.add_argument("xml", type=str, help="""Input xml file""") 
+    parser.add_argument("output", type=str, help="""output text file""") 
+    args = parser.parse_args()
+    return args
+
+def process_xml(xml_handle, output_handle):
+    soup = BeautifulSoup(xml_handle, "xml")
+    for segment in soup.find_all("segment"):
+        who = segment["who"]
+        starttime = segment["starttime"]
+        endtime = segment["endtime"]
+        WMER = segment["WMER"]
+        text = " ".join([element.string for element in segment.find_all("element") if element.string != None])
+        output_handle.write("{} {} {} {} {}\n".format(who, starttime, endtime, WMER, text))
+    xml_handle.close()
+    output_handle.close()
+
+def main():
+    args = get_args()
+
+    xml_handle = open(args.xml, 'r')
+    output_handle = sys.stdout if args.output == '-' else open(args.output, 'w')
+
+    process_xml(xml_handle, output_handle)
+
+if __name__ == "__main__":
+    main()
diff --git a/egs/gale_arabic/s5d/run.sh b/egs/gale_arabic/s5d/run.sh
index 68d15c106dc..f8fdafe0a77 100755
--- a/egs/gale_arabic/s5d/run.sh
+++ b/egs/gale_arabic/s5d/run.sh
@@ -46,6 +46,9 @@ galeData=GALE
 mgb2_dir=""
 giga_dir=""
 
+# preference on how to process xml file (use xml binary or python)
+process_xml=""
+
 run_rnnlm=false
 . ./cmd.sh ## You'll want to change cmd.sh to something that will work on your system.
            ## This relates to the queue.
@@ -64,7 +67,7 @@ if [ $stage -le 0 ]; then
   echo "$0: Preparing data..."
 
   options=""
-  [ ! -z $mgb2_dir ] && options="--mgb2-dir $mgb2_dir"
+  [ ! -z $mgb2_dir ] && options="--process-xml python --mgb2-dir $mgb2_dir"
   local/prepare_data.sh $options
 
   echo "$0: Preparing lexicon and LM..." 
diff --git a/egs/mgb2_arabic/s5/local/check_tools.sh b/egs/mgb2_arabic/s5/local/check_tools.sh
new file mode 100755
index 00000000000..448a6536946
--- /dev/null
+++ b/egs/mgb2_arabic/s5/local/check_tools.sh
@@ -0,0 +1,15 @@
+#!/bin/bash
+
+# check whether bs4 and lxml is installed
+if ! python3 -c "import bs4" 2>/dev/null; then
+  echo "$0: BeautifulSoup4 not installed, you can install it by 'pip install beautifulsoup4' if you prefer to use python to process xml file" 
+  exit 1;
+fi
+
+if ! python3 -c "import lxml" 2>/dev/null; then
+  echo "$0: lxml not installed, you can install it by 'pip install lxml' if you prefer to use python to process xml file"
+  exit 1;
+fi
+
+echo "both BeatufileSoup4 and lxml are installed in python"
+exit 0
diff --git a/egs/mgb2_arabic/s5/local/mgb_data_prep.sh b/egs/mgb2_arabic/s5/local/mgb_data_prep.sh
index afd7f92b7cf..681894a9e29 100755
--- a/egs/mgb2_arabic/s5/local/mgb_data_prep.sh
+++ b/egs/mgb2_arabic/s5/local/mgb_data_prep.sh
@@ -4,8 +4,8 @@
 #               2016-2019  Vimal Manohar
 #               2019 Dongji Gao
 
-if [ $# -ne 2 ]; then
-  echo "Usage: $0 <DB-dir> <mer-sel>"
+if [ $# -ne 3 ]; then
+  echo "Usage: $0 <DB-dir> <mer-sel> <process-xml>"
   exit 1;
 fi
 
@@ -23,12 +23,6 @@ for x in $train_dir $dev_dir; do
   fi
 done
 
-if [ -z $(which xml) ]; then
-  echo "$0: Could not find tool xml"
-  echo "$0: Download and install it from xmlstar.sourceforge.net"
-  exit 1
-fi
-
 find $db_dir/train/wav -type f -name "*.wav" | \
   awk -F/ '{print $NF}' | perl -pe 's/\.wav//g' > \
   $train_dir/wav_list
@@ -39,11 +33,33 @@ head -500 $train_dir/wav_list > $train_dir/wav_list.short
 set -e -o pipefail
 
 xmldir=$db_dir/train/xml/bw
-cat $train_dir/wav_list | while read basename; do
+if [ $process_xml == "python" ]; then
+  echo "using python to process xml file"
+  # check if bs4 and lxml are installin in python
+  local/check_tools.sh
+  # process xml file using python
+  cat $train_dir/wav_list | while read basename; do
     [ ! -e $xmldir/$basename.xml ] && echo "Missing $xmldir/$basename.xml" && exit 1
-    xml sel -t -m '//segments[@annotation_id="transcript_align"]' -m "segment" -n -v  "concat(@who,' ',@starttime,' ',@endtime,' ',@WMER,' ')" -m "element" -v "concat(text(),' ')" $xmldir/$basename.xml | local/add_to_datadir.py $basename $train_dir $mer
-    echo $basename $wavDir/$basename.wav >> $train_dir/wav.scp
-done 
+    local/process_xml.py $xmldir/$basename.xml - | local/add_to_datadir.py $basename $train_dir $mer
+  done
+elif [ $process_xml == 'xml' ]; then
+  # check if xml binary exsits
+  if command -v xml >/dev/null 2>/dev/null; then
+    echo "using xml"
+    cat $train_dir/wav_list | while read basename; do
+      [ ! -e $xmldir/$basename.xml ] && echo "Missing $xmldir/$basename.xml" && exit 1
+      xml sel -t -m '//segments[@annotation_id="transcript_align"]' -m "segment" -n -v  "concat(@who,' ',@starttime,' ',@endtime,' ',@WMER,' ')" -m "element" -v "concat(text(),' ')" $xmldir/$basename.xml | local/add_to_datadir.py $basename $train_dir $mer
+      echo $basename $wavDir/$basename.wav >> $train_dir/wav.scp
+    done
+  else
+    echo "xml not found, you may use python by '--process-xml python'"
+    exit 1;
+  fi
+else
+  # invalid option
+  echo "$0: invalid option for --process-xml, choose from 'xml' or 'python'"
+  exit 1;
+fi
 
 for x in text segments; do
   cp $db_dir/dev/${x}.all $dev_dir/${x}
diff --git a/egs/mgb2_arabic/s5/local/process_xml.py b/egs/mgb2_arabic/s5/local/process_xml.py
new file mode 100755
index 00000000000..3c6eed452ac
--- /dev/null
+++ b/egs/mgb2_arabic/s5/local/process_xml.py
@@ -0,0 +1,35 @@
+#!/usr/bin/env python3
+
+from bs4 import BeautifulSoup
+import sys
+import argparse
+
+def get_args():
+    parser = argparse.ArgumentParser(description="""This script process xml file.""")
+    parser.add_argument("xml", type=str, help="""Input xml file""") 
+    parser.add_argument("output", type=str, help="""output text file""") 
+    args = parser.parse_args()
+    return args
+
+def process_xml(xml_handle, output_handle):
+    soup = BeautifulSoup(xml_handle, "xml")
+    for segment in soup.find_all("segment"):
+        who = segment["who"]
+        starttime = segment["starttime"]
+        endtime = segment["endtime"]
+        WMER = segment["WMER"]
+        text = " ".join([element.string for element in segment.find_all("element") if element.string != None])
+        output_handle.write("{} {} {} {} {}\n".format(who, starttime, endtime, WMER, text))
+    xml_handle.close()
+    output_handle.close()
+
+def main():
+    args = get_args()
+
+    xml_handle = open(args.xml, 'r')
+    output_handle = sys.stdout if args.output == '-' else open(args.output, 'w')
+
+    process_xml(xml_handle, output_handle)
+
+if __name__ == "__main__":
+    main()
diff --git a/egs/mgb2_arabic/s5/run.sh b/egs/mgb2_arabic/s5/run.sh
index a763869903c..e4192c067b0 100755
--- a/egs/mgb2_arabic/s5/run.sh
+++ b/egs/mgb2_arabic/s5/run.sh
@@ -6,6 +6,9 @@
 
 stage=-1
 
+# preference on how to process xml file [python, xml]
+process_xml="python"
+
 . ./cmd.sh
 if [ -f ./path.sh ]; then . ./path.sh; fi
 . utils/parse_options.sh
@@ -50,7 +53,7 @@ fi
 if [ $stage -le 1 ]; then
   #DATA PREPARATION
   echo "Preparing training data"
-  local/mgb_data_prep.sh DB $mer
+  local/mgb_data_prep.sh DB $mer $process_xml
 fi
 
 if [ $stage -le 2 ]; then