NMT-for-CLLS

Application of NMT+Word alignment models for cross-lingual lexical substitution task

Firstly

pip install -r requirements.txt
bash init.sh

Get NMT

• For example

from nmt import NMT_easy


model_name = 'opus-mt' # see https://github.com/UKPLab/EasyNMT

language = 'es'
outpath = 'test.csv'

nmt_model = NMT_easy(model_name)
nmt_semeval2010_2 = nmt_model.make_nmt(<your_df: pd.DataFrame>, language, outpath)

• Or

python3 nmt.py --df_path <path_to_df> --language <your_language> --model_name <nmt_model_name>

Get alignments

• For example

from alignments import Alignments

alignment_models = Alignments()
positions1, spanish_tws1 = alignment_models.model_1(df)
positions2, spanish_tws2 = alignment_models.model_2(df)