North Korean Neural Machine Translation through South Korean Resources

South and North Korea both use the Korean language. However, natural language processing (NLP) research has mostly focused on Therefore, existing NLP systems in language, such as neural machine translation (NMT) systems, cannot properly process inputs. Training a model using data is most straightforward approach to solving this problem, but train NMT models are insufficient. To solve we constructed parallel corpus develop comparable corpus. We manually aligned sentences create evaluation automatically remaining training data. trained our improved quality resources pre-trained model. In addition, propose Korean-specific pre-processing methods, character tokenization, phoneme decomposition more efficiently. demonstrate that consistently improves accuracy compared other methods.