The codes used can be found here.
data_generation/ contains the code used to extract the sentences from the datasets and formatting of dictionary and phoneme mapping.
lstm_models/ contains the code used to train the LSTM model.
mbert_models/ contains the code used to train the multilingual BERT (mBERT) models.
mcnemar_test_mbert/ contains the code used to perform McNemar's Test on the mBERT models.
mono_bert_models/ contains the code used to train monolingual BERT models on monolingual alternatives.
rank_lstm_alts/ contains the code used to rank the scores of each type of alternative (code-switched, English, Chinese) on the LSTM models.
rank_mbert_alts/ contains the code used to rank the scores of each type of alternative (code-switched, English, Chinese) on the mBERT models.