Table 11 F1-score comparison to show the effect of translation and transliteration for sentiment analysis
From: Social media text analytics of Malayalam–English code-mixed using deep learning
Model | Without translation and transliteration | With translation and transliteration |
|---|---|---|
CNN | 0.6778 | 0.7477 |
GRU | 0.6907 | 0.7617 |
LSTM | 0.6794 | 0.7372 |
BiGRU | 0.6659 | 0.7370 |
BiLSTM | 0.6270 | 0.7297 |
LSTM + CNN | 0.6349 | 0.7225 |
GRU + CNN | 0.6663 | 0.7336 |
BiGRU + CNN | 0.6780 | 0.7276 |
BiLSTM + CNN | 0.6736 | 0.7356 |
Transformer based classification model(XLM-R) | 0.6600 | 0.7302 |