Robust Neural Machine Translation with Joint Textual and Phonetic Embedding
This addresses robustness issues in NMT for noisy real-world data, but it is incremental as it focuses on a specific type of noise.
The paper tackles the problem of neural machine translation being sensitive to homophone noises by proposing a method that jointly embeds textual and phonetic information and augments training data with such noises, resulting in significant robustness improvements and even better translation quality on some clean test sets.
Neural machine translation (NMT) is notoriously sensitive to noises, but noises are almost inevitable in practice. One special kind of noise is the homophone noise, where words are replaced by other words with similar pronunciations. We propose to improve the robustness of NMT to homophone noises by 1) jointly embedding both textual and phonetic information of source sentences, and 2) augmenting the training dataset with homophone noises. Interestingly, to achieve better translation quality and more robustness, we found that most (though not all) weights should be put on the phonetic rather than textual information. Experiments show that our method not only significantly improves the robustness of NMT to homophone noises, but also surprisingly improves the translation quality on some clean test sets.