Gene

Linguistics

Philosophy

Vocabulary

Deep neural networks

Machine translation

Messenger RNA

Robustness (evolution)

Regularization (linguistics)

Ambiguity

Biochemistry

Computer science

Translation (biology)

Natural language processing

Segmentation

Artificial intelligence

Chemistry

Programming language

Artificial neural network

Machine learning

Speech recognition

Subword units are an effective way to alleviate the open vocabulary problems in neural machine translation (NMT). While sentences are usually converted into unique subword sequences, subword segmentation is potentially ambiguous and multiple segmentations are possible even with the same vocabulary. The question addressed in this paper is whether it is possible to harness the segmentation ambiguity as a noise to improve the robustness of NMT. We present a simple regularization method, subword regularization, which trains the model with multiple subword segmentations probabilistically sampled during training. In addition, for better subword sampling, we propose a new subword segmentation algorithm based on a unigram language model. We experiment with multiple corpora and report consistent improvements especially on low resource and out-of-domain settings.

Subword Regularization: Improving Neural Network Translation Models with Multiple Subword Candidates