Fully Character-Level Neural Machine Translation without Explicit SegmentationFully Character-Level Neural Machine Translation without Explicit SegmentationJason Lee and Kyunghyun Cho and Thomas Hofmann2016
TLDR; The authors propose a character-level Neural Machine Translation (NMT) architecture. The encoder is a convolutional network with max-pooling and highway layers that reduces size of the source representation. It does not use explicit segmentation. The decoder is a standard RNN. The authors apply their model to WMT'15 DE-EN, CS-EN, FI-EN and RU-EN data in bilingual and multilingual settings. They find that their model is competitive in bilingual settings and significantly outperforms competing models in the multilingual setting with a shared encoder.
#### Key Points
- Challenge: Apply standard seq2seq models to characters is hard because representation is too long. Attention network complexity grows quadratically with sequence length.
- Word-Level models are unable to model rare and out-of-vocab tokens and softmax complexity grows with vocabulary size.
- Character level models are more flexible: No need for explicit segmentation, can model morphological variants, multilingual without increasing model size.
- Reducing the length of the source sentence is key to fast training in char models.
- Encoder Network: Embedding -> Conv -> Maxpool -> Highway -> Bidirectional GRU
- Attenton Network: Single Layer
- Decoder: Two Layer GRU
- Multilingual setting: Language examples are balanced within each batch. No language identifier is provided to the encoder
- Bilingual Results: char2char performs as well as or better than bpe2char or bpe2bpe
- Multilingual Results: char2char outperforms bpe2char
- Trained model is robust to spelling mistakes and unseen morphologies
- Training time: Single Titan X training time for bilingual model is ~2 weeks. ~2.5 updates per second with batch size 64.
- I wonder if you can extract segmentation info from the network post training.
First published: 2016/10/10 (2 years ago) Abstract: Most existing machine translation systems operate at the level of words,
relying on explicit segmentation to extract tokens. We introduce a neural
machine translation (NMT) model that maps a source character sequence to a
target character sequence without any segmentation. We employ a character-level
convolutional network with max-pooling at the encoder to reduce the length of
source representation, allowing the model to be trained at a speed comparable
to subword-level models while capturing local regularities. Our
character-to-character model outperforms a recently proposed baseline with a
subword-level encoder on WMT'15 DE-EN and CS-EN, and gives comparable
performance on FI-EN and RU-EN. We then demonstrate that it is possible to
share a single character-level encoder across multiple languages by training a
model on a many-to-one translation task. In this multilingual setting, the
character-level encoder significantly outperforms the subword-level encoder on
all the language pairs. We also observe that the quality of the multilingual
character-level translation even surpasses the models trained and tuned on one
language pair, namely on CS-EN, FI-EN and RU-EN.