update document

README.md

@@ -0,0 +1,35 @@
+# Multilingual Neural Machine Translation with Soft Decoupled Encoding
+This is the code we used in our paper
+>[Multilingual Neural Machine Translation with Soft Decoupled Encoding](https://arxiv.org/pdf/1902.03499.pdf)
+
+>Xinyi Wang, Hieu Pham, Philip Arthur, Graham Neubig
+
+
+## Requirements
+
+Python 3.6, PyTorch 0.4.1
+
+
+All the scripts for experiments in the paper can be created from the templates under scripts/template/
+
+## Data Processing
+
+The data we use is [multilingual TED corpus](https://github.com/neulab/word-embeddings-for-nmt) by Qi et al.
+
+We provide preprocessed version of the data, which you can get from:
+If you are interested int the details of data processing, you can take a look at the script ``make-eng.sh`` and  ``make-data.sh``.
+
+## Training:
+The template name for the following methods are:
+  1. SDE: bi-semb-bq-o32000
+  2. subword: bi-sw-32000
+  2. subword-joint: bi-sw-joint-32000
+  3. word: bi-w-64000
+
+To make the main experiment scripts for alll 4 languages tested in the paper, simply call
+``bash make-cfg.sh``
+
+## Decoding:
+To make decode scripts, simply use the file make-trans.py. Change the name of the directory where the experiment outputs are stored if you modify the template scripts during training. Otherwise it should just work by calling:
+``python make-trans.py``
+

make-cfg.sh

@@ -1,9 +1,9 @@
 
 TEMP_DIR=scripts/template/
 # change random seed and directory name as desired
-CFG_DIR=cfg_s3/
-VERSION=s3
-SEED=3
+CFG_DIR=cfg_s0/
+VERSION=s0
+SEED=0
 
 mkdir -p scripts/"$CFG_DIR"
 # low-resource language codes
@@ -23,7 +23,7 @@ for i in ${!ILS[*]}; do
   IL=${ILS[$i]}
   RL=${RLS[$i]}
   echo $IL
-  for f in $TEMP_DIR/bi-w-16000 $TEMP_DIR/bi-sw-joint-16000 $TEMP_DIR/bi-sw-16000  $TEMP_DIR/bi-semb-bq-o16000 ; do
+  for f in $TEMP_DIR/bi-w-64000 $TEMP_DIR/bi-sw-joint-32000 $TEMP_DIR/bi-sw-32000  $TEMP_DIR/bi-semb-bq-o32000 ; do
     sed "s/VERSION/$VERSION/g; s/SEED/$SEED/g; s/IL/$IL/g; s/RL/$RL/g" < $f > ${f/template/"$CFG_DIR"/}_$IL$RL.sh 
     chmod u+x ${f/template/"$CFG_DIR"/}_$IL$RL.sh 
   done

make-trans.py

@@ -4,7 +4,7 @@
 
 # make sure the version matches the configuration
 # for example, if in make-cfg.sh, you put s3 as version, there should be an output folder named as outputs_s3/
-version_list = ["s3"]
+version_list = ["s0"]
 temp_dir = { "w": "scripts/template/trans_w", "sw-joint": "scripts/template/trans_sw-joint", "sw": "scripts/template/trans_sw"}
 
 for version in version_list: