Figure 1:
Architecture of SMT system. SMT, statistical machine translation.
Figure 2:
Encoder–decoder model with attention [3].
Figure 3:
Transformer model [21].
Figure 4:
GAN model in the case of NMT use. GAN, generative adversarial network.
Figure 5:
Heat map representation of attention visualization.
Figure 6:
Masked values are represented with zeros.
Figure 7:
Graphical representation of masking operation; the colored right half is the masked part.
Table 1.
For various low-resource corpus SMT outperformed NMT (Ahmadnia et al. [17])
| Corpus | SMT | NMT | NMT* | NMT** |
|---|---|---|---|---|
| Gnome | 20.54 | 15.49 | 17.26 | 18.76 |
| KDE4 | 15.64 | 13.36 | 14.29 | 15.71 |
| Subtitles | 18.82 | 18.62 | 19.51 | 22.54 |
| Ubuntu | 16.76 | 14.27 | 15.14 | 15.87 |
| Tanzil | 17.69 | 15.14 | 16.53 | 17.72 |
| Overall | 17.06 | 15.25 | 16.67 | 17.32 |
[i] SMT, statistical machine translation.
Table 2.
NMT outperformed SMT with transfer learning, ensemble, and further processing of data (Zopth et al.)
| Language | SBMT | NMT | Transfer | Final |
|---|---|---|---|---|
| Hausa | 23.7 | 16.8 | 21.3 | 24.0 |
| Turkish | 20.4 | 11.4 | 17.0 | 18.7 |
| Uzbek | 17.9 | 10.7 | 14.4 | 16.8 |
| Urdu | 17.9 | 5.2 | 13.8 | 14.5 |
[i] SBMT, syntax-based machine translation; SMT, statistical machine translation.
Table 3.
Attention-based NMT outperforms SMT for the Bengali–Hindi language pair (Das et al. [32])
| Translation model | BLEU score | Iterations |
|---|---|---|
| Attention-based NMT model | 20.41 | 25 |
| MOSES (SMT) | 14.35 | - |
[i] SMT, statistical machine translation.
Table 4.
NMT system with transformer model and BPE outperformed phrase-based SMT for English–Hindi and Hindi–English language pairs (Haque et al. [33])
| MT model | BLEU | METEOR | TER |
|---|---|---|---|
| Eng.Hindi-PBSMT | 28.8 | 30.2 | 53.4 |
| Eng.Hindi-NMT | 36.6 | 33.5 | 46.3 |
| Hindi–Eng.PBSMT | 34.1 | 36.6 | 50.0 |
| Hindi–Eng.NMT | 39.9 | 38.5 | 42.0 |
[i] METEOR, Metric for Evaluation of Translation with Explicit Ordering; MT, machine translation; PBSMT, Phrase-Based Statistical Machine Translation System; SMT, statistical machine translation; TER, Translation Error Rate.
Figure 8:
BLEU score generated by NMT and SMT for Eng.–Beng. language pairs. MT, machine translation; SGD, stochastic gradient descent; SMT, statistical machine translation.
Table 5.
English–Bengali translation BLEU scores using different optimizers
| Language pairs | Optimizer | BLEU-4 score | MT model | No. of epochs |
|---|---|---|---|---|
| Eng.–Beng. | Adam | 10.78 | NMT with attention | 12 |
| Eng.–Beng. | SGD | 11.17 | NMT with attention | 12 |
| Eng.–Beng. | 14.58 | MOSES |
[i] MT, machine translation; SGD, stochastic gradient descent.
Figure 9:
BLEU score generated by NMT and SMT for Eng.–Hindi language pairs. MT, machine translation; SGD, stochastic gradient descent; SMT, statistical machine translation.
Table 6.
English–Hindi translation using different optimizers
| Language pair | Optimizer | BLEU-4 score | NMT model | No. of epochs |
|---|---|---|---|---|
| Eng.–Hindi | Adam | 12.25 | NMT with attention | 14 |
| Eng.–Hindi | SGD | 11.50 | NMT with attention | 14 |
| Eng.–Hindi | 16.64 | MOSES |
[i] SGD, stochastic gradient descent.
Figure 10:
BLEU with minimum n-gram having maximum score. SGD, stochastic gradient descent.
Table 7.
BLEU-1, 2, and 3 scores are summarized for Eng.–Beng. and Eng.–Hindi language pairs using Adam- and SGD-Optimizers
| BLEU | Eng.–Beng.-NMT (Adam-Optimizer) | Eng.–Beng.-NMT (SGD-Optimizer) | Eng.–Hindi (NMT-Adam) | Eng.–Hindi (NMT-SGD) |
|---|---|---|---|---|
| BLEU-1 | 14.15 | 13.91 | 15.77 | 14.18 |
| BLEU-2 | 12.65 | 13.11 | 14.12 | 13.33 |
| BLEU-3 | 11.83 | 12.17 | 13.95 | 12.19 |
[i] SGD, stochastic gradient descent.