Construction of Japanese-Chinese Onomatopoeia Corpus Based on Events and Behaviors

Abstract: Numerous manga (Japanese comics) are translated into various languages every year, but the onomatopoeia that depicted in Mangas frequently controls the strength and nuance of the impression by repeatedly using prompts and long notes often remains untranslated, which prevents many manga fans from having the full reading experience. This study focuses on translating Japanese onomatopoeia into Chinese as a first step. The ultimate goal for this research is to translate onomatopoeia into multiple languages. In previous research, the focus was primarily on translating the onomatopoeia itself, whereas this study proposes a method that translates both the onomatopoeia and its related words. Toward this end, we constructed a Japanese-Chinese onomatopoeia (JCO) corpus. Our dataset includes an accurate translation of each onomatopoeia as well as a related word that makes the sound or indicates the event or behavior. The corpus is mainly composed of onomatopoeia found in Manga 109, with additional entries collected from the internet. The most appropriate word is determined by calculating the cosine similarity and Levenshtein distance. If the word cannot be determined from our dataset, we use the International Phonetic Alphabet (IPA) to transcribe the Japanese characters into their closest Chinese phonetic counterparts. The onomatopoeia depicted in mangas frequently controls the strength and nuance of the impression by repeatedly using prompts and long notes, which adds to the complexity of translating these terms accurately. This is because the repeated use of certain characters and the lengthening of sounds in onomatopoeia can convey different intensities and subtleties of the depicted actions or emotions. We also developed a tool to aid manga translators in determining the most appropriate translation for onomatopoeia. The tool works by allowing users to input Japanese onomatopoeia and their related words, and then automatically outputs the most suitable Chinese translation.During our experiments, we randomly selected five manga from Manga 109 and extracted 100 unique onomatopoeic words for testing. The resulting translations were evaluated by native Chinese speakers who are also proficient in Japanese and Japanese speakers who are also proficient in Chinese, and their overall feedback was positive, indicating that the corpus and tool could be used to translate onomatopoeia with sufficient accuracy.

Keywords: Japanese-Chinese corpus, manga, translation of onomatopoeia, translation with related word

DOI: 10.54941/ahfe1005578

Cite this paper:

​