English to Sinhala Neural Machine Translation

• Project Mentor
  1. Dr Uthayasanker Thayasivam
• Contributors
  1. Thilakshi Fonseka
  2. Rashmini Naranpanawa
  3. Ravinga Perera

---

Summary

This research is about developing a NMT system using Transformer architecture for the under-resourced, domain-specific English to Sinhala translation task. The translation quality is improved by exploring effective ways of incorporating Part-of-Speech (POS) information and subword techniques.

Description

This project consists of the following.

• Transformer baseline
• Transformer with subword segmentation
  • Byte Pair Encoding
  • Unigram based subword regularization
• Transformer with Part-of-Speech (POS)
  • Input embedding
  • Positional encoding

Architecture Diagrams

Following are the architecture diagrams for the POS integration with the input embedding and positional encoding respectively.

The following instructions will guide to produce our results.

Requirements

We use fairseq for training, sentencepiece for preprocessing & sacrebleu to produce BLEU scores.

Transformer Baseline

pip install fairseq sacrebleu 

Transformer with subword segmentation

pip install fairseq sacrebleu sentencepiece

Transformer with POS

pip install sacrebleu sentencepiece

Since POS is implemented withing the fairseq-transformer, navigate to the project directory and install fairseq as following

pip install --editable ./

Train the baseline transformer model

• Navigate to src/Transformer-baseline. Follow the instructions given in the README.md.

Train the subword segmented transformer models

• To train the Transformer BPE model, navigate to src/Subword-segmentation/Transformer-BPE. Follow the instructions given in the README.md.
• To train the Transformer subword regularization model, navigate to src/Subword-segmentation/Transformer-subword-regularization. Follow the instructions given in the README.md.

Train the POS integrated transformer models

• Navigate to src/POS-implementation. Follow the instructions given in the README.md.

---

Publications

T. Fonseka, R. Naranpanawa, R. Perera and U. Thayasivam, "English to Sinhala Neural Machine Translation," 2020 International Conference on Asian Language Processing (IALP), Kuala Lumpur, Malaysia, 2020, pp. 305-309, doi: 10.1109/IALP51396.2020.9310462.

R. Naranpanawa, R. Perera, T. Fonseka and U. Thayasivam, "Analyzing Subword Techniques to Improve English to Sinhala Neural Machine Translation," International Journal of Asian Language Processing (IJALP), vol. 30, no. 04, p. 2050017, 2020, doi: 10.1142/s2717554520500174.

License

Apache License 2.0

Code of Conduct

Please read our code of conduct document here.