An Evaluation of Neural Machine Translation Models on Historical Spelling Normalization
This work addresses the problem of normalizing historical spellings for researchers and linguists working with multiple languages, but it is incremental as it applies existing NMT methods to a specific domain.
The paper tackled historical spelling normalization for five languages using various neural machine translation models, finding that NMT models outperform SMT models in character error rate and that subword-level models with small vocabularies are better for low-resource languages, with a hybrid method proposed to further improve performance.
In this paper, we apply different NMT models to the problem of historical spelling normalization for five languages: English, German, Hungarian, Icelandic, and Swedish. The NMT models are at different levels, have different attention mechanisms, and different neural network architectures. Our results show that NMT models are much better than SMT models in terms of character error rate. The vanilla RNNs are competitive to GRUs/LSTMs in historical spelling normalization. Transformer models perform better only when provided with more training data. We also find that subword-level models with a small subword vocabulary are better than character-level models for low-resource languages. In addition, we propose a hybrid method which further improves the performance of historical spelling normalization.