This repo contains the working code for the models submitted in the Hope Speech Detection in English-Malayalam code-mixed shared task in LT-EDI-EACL2021 . We achieved a team rank of 9(with 0.73 f1-score) of the 31 participating teams.
-
BiLSTM with attention layer both word2vec and fasttext word embeddedings were used
-
CNN - A 1D CNN model
-
Optuna is used for hyperparameter tuning