Character-level Chinese-English Translation through ASCII Encoding

Proceedings of the Third Conference on Machine Translation (WMT), Volume 1: Research Papers, pages 10–16 Belgium, Brussels, October 31 - Novermber 1, 2018. c⃝2018 Association for Computational Linguistics https://doi.org/10.18653/v1/W18-64002 Character-level Chinese-English Translation through ASCII Encoding Nikola I. Nikolov∗, Yuhuang Hu∗, Mi Xue Tan, Richard H.R. Hahnloser Institute of Neuroinformatics, University of Z¨urich and ETH Z¨urich, Switzerland {niniko, yuhuang.hu, mtan, rich}@ini.ethz.ch Abstract Character-level Neural Machine Translation (NMT) models have recently achieved impres- sive results on many language pairs. They mainly do well for Indo-European language pairs, where the languages share the same writing system. However, for translating be- tween Chinese and English, the gap between the two different writing systems poses a ma- jor challenge because of a lack of system- atic correspondence between the individual linguistic units. In this paper, we enable character-level NMT for Chinese, by breaking down Chinese characters into linguistic units similar to that of Indo-European languages. We use the Wubi encoding scheme1, which preserves the original shape and semantic in- formation of the characters, while also being reversible. We show promising results from training Wubi-based models on the character- and subword-level with recurrent as well as convolutional models. 1 Introduction Character-level sequence-to-sequence (Seq2Seq) models for machine translation can perform comparably to subword-to-subword or subword- to-character models, when dealing with Indo- European language pairs, such as German-English or Czech-English (Lee et al., 2017). Such lan- guage pairs benefit from having a common Latin character representation, which facilitates suitable character-to-character mappings to be learned. This method, however, is more difficult for non- Latin language pairs, such as Chinese-English. Chinese characters differ from English characters, in the sense that they carry more meaning and resemble subword units in English. For exam- ple, the Chinese character ‘人’ corresponds to the 1Code and data available at https://github.com/ duguyue100/wmt-en2wubi. * Equal contribution ������� b d y a d bd|yad Seq2Seq commitment ……… ……… ……… ……… Figure 1: Overview of the wubi2en approach to Chinese-to-English translation. A raw Chinese word (‘承诺’) is encoded into ASCII characters (‘bd|yad’), using the Wubi encoding method, be- fore passing it to a Seq2Seq network. The net- work generates the English translation ‘commit- ment’, processing one ASCII character at a time. word ‘human’ in English. This lack of correspon- dence makes the problem more demanding for a Chinese-English character-to-character model, as it would be forced to map higher-level linguis- tic units in Chinese to individual Latin characters in English. Good performance on this task may, therefore, require specific architectural decisions. In this paper, we propose a simple solution to this challenge: encode Chinese into a meaning- ful string of ASCII characters, using the Wubi method (Lunde, 2009) (Section 3). This encoding enables efficient and accurate character-level pre- diction applications in Chinese, with no changes required to the model architecture (see Figure 1). Our approach significantly reduces the character vocabulary size of a Chinese text, while preserv- ing the shape and semantic information encoded in the Chinese characters. 10 We demonstrate the utility of the Wubi en- coding on …

Content truncated to 500 words