This repo contains a dockerfile and instructions to translate a text using an statistical dictionary.
To build the image, you just need to run the following command (provided that you are inside this repo's directory. Otherwise replace . with the proper path):
docker build -t sd .
Prior to using the dictionary, you need to train the alignment model. You can do so running the following command:
docker container run -it --rm \
-v "$(pwd)"/source_file:/sd/source_file \
-v "$(pwd)"/target_file:/sd/target_file \
-v "$(pwd)"/dest_dir:/sd/dest_dir \
sd scr/IBM1Alignments.sh source_file target_file dest_dir
where source_file is the source from the parallel training data; target_file is the target from the parallel training data; and dest_dir is the directory in which to save the alignments.
Finally, you can translate a document by doing:
docker container run -it --rm \
-v "$(pwd)"/text_file:/sd/text_file \
-v "$(pwd)"/dest_dir/alignments:/sd/alignments \
sd python scr/SD.py -t text_file -a alignments > hyp
where text_file is the text to translate; alignments is the file created at the alignments step (dest_dir/alignmetns); and hyp is the resulting translation.