earticle

영어

With a keen interest in the quality of MT output, errors in raw MT output have received much attention in domestic translation studies. However, errors in post-edited output have rarely been discussed, although a considerable amount of errors in raw MT output remain uncorrected even after the post-editing task. Against this backdrop, this study aims to investigate errors in post-edited output with Korean-English parallel translation corpus for AI training, released in 2019 by the National Information Society Agency. For this purpose, 300 parallel sentences with errors were collected in economic news corpus and then classified by error type. Analysis results showed a wide array of errors ranging from accuracy to readability, indicating the need to examine errors in post-edited output and major factors affecting the post-editing process.