diff --git a/lib/src/main/scala/com/worksap/nlp/uzushio/lib/runners/DeduplicateParagraphs.scala b/lib/src/main/scala/com/worksap/nlp/uzushio/lib/runners/DeduplicateParagraphs.scala
index 52896bf..5346df2 100644
--- a/lib/src/main/scala/com/worksap/nlp/uzushio/lib/runners/DeduplicateParagraphs.scala
+++ b/lib/src/main/scala/com/worksap/nlp/uzushio/lib/runners/DeduplicateParagraphs.scala
@@ -820,7 +820,7 @@ object DeduplicateParagraphs {
 
     val basicCols = (if (debug) {
                        joined.columns.filter {
-                         case "parHash" => false
+                         case "parHash" => true
                          case "exactFreq" | "nearFreq" => false
                          case _ => true
                        }
@@ -901,7 +901,7 @@ object DeduplicateParagraphs {
 
   case class ProcessedDocument(text: String, filter: String)
 
-  private def processDocumentParts(
+    private def processDocumentParts(
       args: Args,
       docId: String,
       parts: IndexedSeq[Paragraph]
diff --git a/scripts/submit_all_compute_stats.sh b/scripts/submit_all_compute_stats.sh
new file mode 100644
index 0000000..1a77b0d
--- /dev/null
+++ b/scripts/submit_all_compute_stats.sh
@@ -0,0 +1,70 @@
+submit_post2017() {
+    qsub -g gcf51199 -l rt_F=10 -l h_rt=4:00:00 submit_dedup_stage1.sh \
+        "/groups/gcf51199/cc/extracted/segment\=$1" \
+        /groups/gcf51199/cc/stats_raw_v2/segment=$1 \
+        500 4000
+}
+
+submit_post2017 CC-MAIN-2017-04
+submit_post2017 CC-MAIN-2017-09
+submit_post2017 CC-MAIN-2017-13
+submit_post2017 CC-MAIN-2017-17
+submit_post2017 CC-MAIN-2017-22
+submit_post2017 CC-MAIN-2017-26
+submit_post2017 CC-MAIN-2017-30
+submit_post2017 CC-MAIN-2017-34
+submit_post2017 CC-MAIN-2017-39
+submit_post2017 CC-MAIN-2017-43
+submit_post2017 CC-MAIN-2017-47
+submit_post2017 CC-MAIN-2017-51
+submit_post2017 CC-MAIN-2018-05
+submit_post2017 CC-MAIN-2018-09
+submit_post2017 CC-MAIN-2018-13
+submit_post2017 CC-MAIN-2018-17
+submit_post2017 CC-MAIN-2018-22
+submit_post2017 CC-MAIN-2018-26
+submit_post2017 CC-MAIN-2018-30
+submit_post2017 CC-MAIN-2018-34
+submit_post2017 CC-MAIN-2018-39
+submit_post2017 CC-MAIN-2018-43
+submit_post2017 CC-MAIN-2018-47
+submit_post2017 CC-MAIN-2018-51
+submit_post2017 CC-MAIN-2019-04
+submit_post2017 CC-MAIN-2019-09
+submit_post2017 CC-MAIN-2019-13
+submit_post2017 CC-MAIN-2019-18
+submit_post2017 CC-MAIN-2019-22
+submit_post2017 CC-MAIN-2019-26
+submit_post2017 CC-MAIN-2019-30
+submit_post2017 CC-MAIN-2019-35
+submit_post2017 CC-MAIN-2019-39
+submit_post2017 CC-MAIN-2019-43
+submit_post2017 CC-MAIN-2019-47
+submit_post2017 CC-MAIN-2019-51
+submit_post2017 CC-MAIN-2020-05
+submit_post2017 CC-MAIN-2020-10
+submit_post2017 CC-MAIN-2020-16
+submit_post2017 CC-MAIN-2020-24
+submit_post2017 CC-MAIN-2020-29
+submit_post2017 CC-MAIN-2020-34
+submit_post2017 CC-MAIN-2020-40
+submit_post2017 CC-MAIN-2020-45
+submit_post2017 CC-MAIN-2020-50
+submit_post2017 CC-MAIN-2021-04
+submit_post2017 CC-MAIN-2021-10
+submit_post2017 CC-MAIN-2021-17
+submit_post2017 CC-MAIN-2021-21
+submit_post2017 CC-MAIN-2021-25
+submit_post2017 CC-MAIN-2021-31
+submit_post2017 CC-MAIN-2021-39
+submit_post2017 CC-MAIN-2021-43
+submit_post2017 CC-MAIN-2021-49
+submit_post2017 CC-MAIN-2022-05
+submit_post2017 CC-MAIN-2022-21
+submit_post2017 CC-MAIN-2022-27
+submit_post2017 CC-MAIN-2022-33
+submit_post2017 CC-MAIN-2022-40
+submit_post2017 CC-MAIN-2022-49
+submit_post2017 CC-MAIN-2023-06
+submit_post2017 CC-MAIN-2023-14
+submit_post2017 CC-MAIN-2023-23
\ No newline at end of file
diff --git a/scripts/submit_all_compute_stats_old.sh b/scripts/submit_all_compute_stats_old.sh
new file mode 100644
index 0000000..14386f6
--- /dev/null
+++ b/scripts/submit_all_compute_stats_old.sh
@@ -0,0 +1,11 @@
+submit_pre2016() {
+    qsub -g gcf51199 -l rt_F=10 -l h_rt=4:00:00 submit_dedup_stage1.sh \
+        "/groups/gcf51199/cc2/extracted/$1" \
+        "/groups/gcf51199/cc/stats_raw_v2/segment=$1" \
+        500 4000
+}
+
+submit_pre2016 merged-2013
+submit_pre2016 merged-2014
+submit_pre2016 merged-2015
+submit_pre2016 merged-2016
\ No newline at end of file
diff --git a/scripts/submit_all_filter.sh b/scripts/submit_all_filter.sh
index 5f6a246..288ccfa 100644
--- a/scripts/submit_all_filter.sh
+++ b/scripts/submit_all_filter.sh
@@ -4,7 +4,7 @@ submit() {
     qsub -g gcf51199 -l rt_F=10 -l h_rt=1:00:00 submit_filter_debug_2.sh \
         "/groups/gcf51199/cc/extracted/segment\=$1" \
         /groups/gcf51199/cc/stats_merged_v1/for_filter/all \
-        "/groups/gcf51199/cc/filtered_v1/segment=$1"
+        "/groups/gcf51199/cc/filtered_v2/segment=$1"
 }
 
 submit CC-MAIN-2017-04
@@ -23,7 +23,7 @@ submit CC-MAIN-2018-05
 submit CC-MAIN-2018-09
 submit CC-MAIN-2018-13
 submit CC-MAIN-2018-17
-# submit CC-MAIN-2018-22
+submit CC-MAIN-2018-22
 submit CC-MAIN-2018-26
 submit CC-MAIN-2018-30
 submit CC-MAIN-2018-34
diff --git a/scripts/submit_dedup_stage1.sh b/scripts/submit_dedup_stage1.sh
index 0f9eafb..20794de 100644
--- a/scripts/submit_dedup_stage1.sh
+++ b/scripts/submit_dedup_stage1.sh
@@ -3,6 +3,7 @@
 #$ -j y
 #$ -cwd
 #$ -l USE_SSH=1
+#$ -l USE_EXTRA_NETWORK=1
 
 du -hs "$1" > /dev/null &
 
@@ -38,6 +39,7 @@ NUM_PARTITIONS_PROPAGATION=${4:-$(($NUM_PARTITIONS * 4))}
     --conf spark.eventLog.dir=/scratch/$USER/spark-exlog \
     --conf spark.local.dir=$SPARK_LOCAL_DIRS \
     --conf spark.sql.shuffle.partitions=${NUM_PARTITIONS_PROPAGATION} \
+    --conf spark.sql.parquet.columnarReaderBatchSize=512 \
     local://$UZUSHIO_JAR \
     --input="$INPUT" \
     --output="$OUTPUT" \
diff --git a/scripts/submit_filter_debug_2.sh b/scripts/submit_filter_debug_2.sh
index 04bdb69..44b9da8 100644
--- a/scripts/submit_filter_debug_2.sh
+++ b/scripts/submit_filter_debug_2.sh
@@ -50,7 +50,7 @@ NUM_PARTITIONS_PROPAGATION=4000
     --filters=$SCRIPT_DIR/pipeline_02.conf \
     --partitions=$NUM_PARTITIONS \
     --execution=filter-debug \
-    -Pkenlm=/groups/gcf51199/filter/n-gram_model/kenlm_model_sudachi_filter.bin \
+    -Pkenlm=/groups/gcf51199/filter/n-gram_model/kenlm_merge-code_0.05_model.bin \
     -Psudachi=/groups/gcf51199/resources/sudachi-dictionary-20230927/system_core.dic \
     --format=json --compression=gzip --text-only