Improving a Multi-Source Neural Machine Translation Model with Corpus Extension for Low-Resource Languages