Transfer Learning for Low-Resource Neural Machine Translation
Abstrak
The encoder-decoder framework for neural machine translation (NMT) has been shown effective in large data scenarios, but is much less effective for low-resource languages. We present a transfer learning method that significantly improves Bleu scores across a range of low-resource languages. Our key idea is to first train a high-resource language pair (the parent model), then transfer some of the learned parameters to the low-resource pair (the child model) to initialize and constrain training. Using our transfer learning method we improve baseline NMT models by an average of 5.6 Bleu on four low-resource language pairs. Ensembling and unknown word replacement add another 2 Bleu which brings the NMT performance on low-resource machine translation close to a strong syntax based machine translation (SBMT) system, exceeding its performance on one language pair. Additionally, using the transfer learning model for re-scoring, we can improve the SBMT system by an average of 1.3 Bleu, improving the state-of-the-art on low-resource machine translation.
Topik & Kata Kunci
Penulis (4)
Barret Zoph
Deniz Yuret
Jonathan May
Kevin Knight
Akses Cepat
- Tahun Terbit
- 2016
- Bahasa
- en
- Total Sitasi
- 905×
- Sumber Database
- Semantic Scholar
- DOI
- 10.18653/v1/D16-1163
- Akses
- Open Access ✓