Have a personal or library account? Click to login
Performance Comparison of Statistical vs. Neural-Based Translation System on Low-Resource Languages Cover

Performance Comparison of Statistical vs. Neural-Based Translation System on Low-Resource Languages

International Journal on Smart Sensing and Intelligent Systems
Volume 16 (2023): Issue 1 (January 2023)
By: Goutam Datta,  Nisheeth Joshi and  Kusum Gupta  
Open Access
|Aug 2023

Figures & Tables

Figure 1:

Architecture of SMT system. SMT, statistical machine translation.
Architecture of SMT system. SMT, statistical machine translation.

Figure 2:

Encoder–decoder model with attention [3].
Encoder–decoder model with attention [3].

Figure 3:

Transformer model [21].
Transformer model [21].

Figure 4:

GAN model in the case of NMT use. GAN, generative adversarial network.
GAN model in the case of NMT use. GAN, generative adversarial network.

Figure 5:

Heat map representation of attention visualization.
Heat map representation of attention visualization.

Figure 6:

Masked values are represented with zeros.
Masked values are represented with zeros.

Figure 7:

Graphical representation of masking operation; the colored right half is the masked part.
Graphical representation of masking operation; the colored right half is the masked part.

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.
BLEU score generated by NMT and SMT for Eng.–Beng. language pairs. MT, machine translation; SGD, stochastic gradient descent; SMT, statistical machine translation.

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.
BLEU score generated by NMT and SMT for Eng.–Hindi language pairs. MT, machine translation; SGD, stochastic gradient descent; SMT, statistical machine translation.

Figure 10:

BLEU with minimum n-gram having maximum score. SGD, stochastic gradient descent.
BLEU with minimum n-gram having maximum score. SGD, stochastic gradient descent.

Attention-based NMT outperforms SMT for the Bengali–Hindi language pair (Das et al_ [32])

Translation modelBLEU scoreIterations
Attention-based NMT model20.4125
MOSES (SMT)14.35-

NMT outperformed SMT with transfer learning, ensemble, and further processing of data (Zopth et al_)

LanguageSBMTNMTTransferFinal
Hausa23.716.821.324.0
Turkish20.411.417.018.7
Uzbek17.910.714.416.8
Urdu17.95.213.814.5

NMT system with transformer model and BPE outperformed phrase-based SMT for English–Hindi and Hindi–English language pairs (Haque et al_ [33])

MT modelBLEUMETEORTER
Eng.Hindi-PBSMT28.830.253.4
Eng.Hindi-NMT36.633.546.3
Hindi–Eng.PBSMT34.136.650.0
Hindi–Eng.NMT39.938.542.0

English–Hindi translation using different optimizers

Language pairOptimizerBLEU-4 scoreNMT modelNo. of epochs
Eng.–HindiAdam12.25NMT with attention14
Eng.–HindiSGD11.50NMT with attention14
Eng.–Hindi 16.64MOSES

English–Bengali translation BLEU scores using different optimizers

Language pairsOptimizerBLEU-4 scoreMT modelNo. of epochs
Eng.–Beng.Adam10.78NMT with attention12
Eng.–Beng.SGD11.17NMT with attention12
Eng.–Beng. 14.58MOSES

BLEU-1, 2, and 3 scores are summarized for Eng_–Beng_ and Eng_–Hindi language pairs using Adam- and SGD-Optimizers

BLEUEng.–Beng.-NMT (Adam-Optimizer)Eng.–Beng.-NMT (SGD-Optimizer)Eng.–Hindi (NMT-Adam)Eng.–Hindi (NMT-SGD)
BLEU-114.1513.9115.7714.18
BLEU-212.6513.1114.1213.33
BLEU-311.8312.1713.9512.19

For various low-resource corpus SMT outperformed NMT (Ahmadnia et al_ [17])

CorpusSMTNMTNMT*NMT**
Gnome20.5415.4917.2618.76
KDE415.6413.3614.2915.71
Subtitles18.8218.6219.5122.54
Ubuntu16.7614.2715.1415.87
Tanzil17.6915.1416.5317.72
Overall17.0615.2516.6717.32
DOI: https://doi.org/10.2478/ijssis-2023-0007 | Journal eISSN: 1178-5608
Journal RSS Feed
Language: English
Submitted on: Feb 22, 2023
|
Published on: Aug 12, 2023
Published by: Professor Subhas Chandra Mukhopadhyay
In partnership with: Paradigm Publishing Services
Publication frequency: 1 issue per year
Keywords:
Neural machine translation,
statistical machine translation,
RNN,
deep learning
Related subjects:
Engineering,
Introductions and overviews,
Engineering, other

© 2023 Goutam Datta, Nisheeth Joshi, Kusum Gupta, published by Professor Subhas Chandra Mukhopadhyay
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.

Previous articleVolume 16 (2023): Issue 1 (January 2023)Next article