diff --git a/bin/split-addresses.sh b/bin/split-addresses.sh
index 933dbbe..72d0ca8 100755
--- a/bin/split-addresses.sh
+++ b/bin/split-addresses.sh
@@ -2,5 +2,6 @@
 
 for area in $(cat data/areas-train.txt)
 do
-    grep -E "^${area}\s" data/area-address-train.tsv > "areas/${area}/addresses-train.txt"
+    grep -E "^${area}\s" data/area-address-train.tsv | \
+    cut -f2 > "areas/${area}/addresses-train.txt"
 done
diff --git a/dvc.lock b/dvc.lock
index 9240cdc..6bb5b0a 100644
--- a/dvc.lock
+++ b/dvc.lock
@@ -58,8 +58,8 @@
       md5: f8363cdb0c36496eac9fcb89f367bffd
       size: 305
     - path: bin/split-addresses.sh
-      md5: a673fe80ebde2c236ede30c4ed57ce8c
-      size: 160
+      md5: e64b86f9e4c88ef778e19a7423e59b0b
+      size: 176
     - path: data/area-address-train.tsv
       md5: 179a8eb09f2ddf72dd9aa2add136fb87
       size: 4010846
@@ -68,8 +68,8 @@
       size: 964
     outs:
     - path: areas
-      md5: 880febe8c0d19ec9ae40a29f5bf026ba.dir
-      size: 4010846
+      md5: 9be0cb2ec76928e1443e2d4679c014c3.dir
+      size: 3694627
       nfiles: 48
   prepare@train:
     cmd: node bin/address-to-area.mjs  "data/addresses-40-cnrs-rnsr-big-classes-train.txt"