Earticle

This study aims to identify stylistic differences between human and machine translation in terms of word usage. For this purpose, a comparable corpus was constructed, which consisted of English translations of 110 Korean newspaper editorials done by a group of human translators and three online machine translation services (Google, Bing and Papago) respectively. Principal component analysis was performed on the corpus to investigate differences in the way 200 most frequent terms are related to the human and machine translators. Additionally, part-of-speech analyses were carried out to further elucidate the differences found in the PCA analysis. The major findings are that machine translators tend to overuse ‘be’ verbs and ‘as’ and ‘if’ subjunctive connectives, while underusing third-person personal pronouns, particularly the female form, ‘she’. Additionally, they were found to rely heavily on high-frequency content words. These characteristics of machine translators are construed as stemming from their scope of lexical options being limited by structural correspondence to original Korean texts.

학문 간의 넘나듦과 소통이 무엇보다 중요한 통섭의 시대에 전산 문체론과 저자 판별이라는 분야는 언어학과 전산학, 문학의 통합적인 접근이 요구된다는 점에서 최근 많은 관심을 받고 있다. 이 연구에서는 초기 현대소설 70편(14명의 저자)을 대상으로 하여 언어 사용 양상을 계량적으로 살펴보고 통계적인 방법을 이용하여 작가별 문체의 특성을 규명해 보았다. 특히 저자별 문체의 특성을 밝히기 위해 문장 길이를 포함하여 일반명사, 동사, 형용사, 부사와 같은 어휘 범주, 조사, 어미, 보조용언과 같은 기능 범주들의 사용 양상을 t-점수를 토대로 고찰하고 작가 간 문체상의 유사도를 시각화하여 제시하였다. 그 결과로 우리는 각 언어 특성이 저자의 문체적 특성을 밝히는 데 유효함을 보일 수 있었으며 그 가운데서도 어휘 범주에 비해 폐쇄적이고 저자의 의도적인 선택이 덜 개입된 기능 범주의 사용 특성이 저자 판별에 더욱 유용할 것임을 제안하였다.

Recently interest in computational stylistics and authorship attribution has been growing through cooperation between the fields of linguistics, literature, and computer science. This study aims to quantitatively investigate the aspects of language use and to capture the stylistic characteristics of 14 authors using the statistical approach based on 70 early modern Korean novels. Specifically, to clarify the stylistic characteristics of each author, we consider eight linguistic features including sentence length, lexical categories (common nouns, verbs, adjectives, and adverbs), functional categories (particles, endings, and auxiliary verbs) and visualize the novels in a 2-dimensional space according to the similarities between them. Our result show that the analysis of these linguistic features is an effective method for capturing the stylistic characteristics of the authors. We propose that a usage analysis of linguistic elements belonging to functional categories is more useful when it comes to authorship attribution an analysis of lexical categories.

This research defines the uniqueness that is categorized in writing as style and aims to establish the special characteristic of style through a quantitative research method. Research on style is an area of interest that is popular in both linguistics and literature, however it is not fully researched in any area of academia. Research on style in Korea generally until now only used qualitative research methods only and is limited on a more objective and quantitative methods. Therefore, this paper specifically shows objective evidence that a quantitative research method founded on computerization is very useful in researching style. Through this we attempted to suggest the function of computerized style. This research took 34 novels by Chae Mansik, Kim Nam Chun, and Lee Ki Young and quantitatively analyzed the uniqueness of the language and explored the stylistic characteristics of these authors. As a result, we were able to see the characteristics differences quantitatively the novels by Chae Mansik, Kim Nam Chun, and Lee Ki Young possessed. Through the differences of usage of these language characteristics, we were able to explain the three author's stylistic characteristics. The method suggested by the paper utilizes together the original qualitative style research method can be applied to investigate a specific author’s style characteristic. This is significant as we verified the functionality of the computerized style method.

이 연구는 한국 현대소설을 대표하는 주요 작가 30명의 소설, 총 363편을 대상으로 계량적 문체 분석을 수행하고 이를 통해 한국 현대소설의 문체적 계보를 수립할 수 있는 단서를 모색해 보는 데 목적이 있다. 개인의 문체에 대한 연구는 국내에서는 주로 정성적인 접근이 많이 이루어진 것에 비해 국외에서는 다양한 정량적 접근이 문체 분석, 저자 판별 등에 활용되어 왔다. 본고에서는 국내외에서 수행된 계량적 문체 분석 방법 가운데 품사별 어휘 사용 빈도, 특히 기능범주의 사용 양상을 검토하였고, 다른 한편으로는 저자별, 작품별 유사도를 코사인 유사도를 통해 측정함으로써 저자들의 문체적 특성을 규명하고자 하였다. 특히 품사별 언어 사용에서는 일반부사, 접속부사, 부사격조사, 연결어미 등이 주로 논의되었으며 이 과정에서 기능범주의 사용이 저자의 특성을 드러내는 주요한 자질이 되고 있음을 보였다. 또한 저자별, 작품별 유사도를 코사인 유사도로 측정하고 이를 토대로 저자들 사이의 문체적 거리뿐 아니라 동일 저자가 저술한 작품들 사이의 차이도 가늠해 볼 수 있었다. 마지막으로 개별 저자의 문체적 특성이 온전히 규명되기 위해서는 본 연구에서 제시한 방법론에 더하여 개별 언어 요소가 문체에 미치는 영향 등에 대한 더욱 심층적인 분석이 필요함을 제안하였다.

The aim of this study is to conduct a quantitative stylistic analysis of the novels of 30 major authors representing Korean modern literature, totaling 363 works, and to seek clues to establish the stylistic genealogy of Korean modern literature. While research on individual styles in South Korea has mainly taken a qualitative approach, various quantitative methods have been utilized abroad for stylistic analysis and authorship attribution. In this paper, among the quantitative stylistic analysis methods conducted domestically and internationally, we examined the frequency of vocabulary usage by part of speech, particularly the patterns of functional categories. On the other hand, we aimed to identify the stylistic characteristics of the authors by measuring the similarity between authors and works using cosine similarity. In terms of language use by part of speech, general adverbs, conjunctive adverbs, adverbial markers, and linking endings were primarily discussed, demonstrating that the use of functional categories is a significant trait that reveals the author's characteristics. Furthermore, by measuring the similarity between authors and works using cosine similarity, we could gauge the stylistic distance between authors as well as the differences between works written by the same author. Finally, it was suggested that to fully elucidate the stylistic characteristics of individual authors, a more in-depth analysis of the impact of individual linguistic elements on style, in addition to the methodology presented in this study, is necessary.