In these two notebooks, I implement Neural Machine Translation (NMT) using Long Short-Term Memory (LSTM) with teacher forcing and Transformer models.
After running several experiments, I calculated the BLEU scores for both models. The results show that the Transformer model outperforms the LSTM model in terms of BLEU scores.
Below is a table summarizing the average BLEU scores for both models:
| Model | Average BLEU score | Average 1-gram score |
|---|---|---|
| LSTM | 0.0023 | 0.0101 |
| Transformer | 0.0902 | 0.4134 |
As shown in Table1., the Transformer model achieved a significantly higher average BLEU score than the LSTM model. This indicates that the Transformer model is more effective in accurately translating text from one language to another.