AKS-DHLAB/KPoEM

KPoEM (Korean Poetry Emotion Mapping) Dataset

We collected 483 poems from five prominent Korean modern poets: Kim So-wol (165 poems), Yun Dong-ju (113 poems), Im Hwa (44 poems), Yi Sang (47 poems), and Han Yong-un (118 poems). We scraped their major works from public domain sources, including "Azaleas" (Korean: 진달래꽃, Jindallaekkot), "Sky, Wind, Stars, and Poetry" (Korean: 하늘과 바람과 별과 시, Haneulgwa baramgwa byeolgwa si), "Hyunhaetan" (Korean: 현해탄, Hyeonhaetan; Hanja: 玄海灘) , "Silence of My Beloved" (Korean: 님의 침묵, Nimui chimmuk), In the case of Yi Sang, the dataset focuses on his Korean-language series poems such as "Crow's Eye View" (Korean: 오감도, Ogamdo; Hanja: 烏瞰圖), "Reverse" (Korean: 역단, Yeokdan; Hanja: 易斷), and "Critical Condition" (Korean: 위독, Widok; Hanja: 危篤).

In results, this dataset has a total of 7,007 emotion-annotated line-level text.

In v4, the names of some columns that were originally in Korean were changed to English.

• Prerprint Paper: Decoding the Poetic Language of Emotion in Korean Modern Poetry: Insights from a Human-Labeled Dataset and AI Modeling

KPoEM 데이터셋 사용 안내

본 저장소에는 두 가지 데이터셋이 포함되어 있습니다.

• KPoEM_line_dataset_v4.tsv : 시 행 단위(line-level) 데이터셋

• KPoEM_poem_dataset_v4.tsv : 시 작품 단위(poem-level) 데이터셋

⚠️ 주의: 두 데이터셋의 컬럼 구조가 달라서 datasets.load_dataset()으로 동시에 불러올 경우 CastError가 발생합니다. 따라서 개별 파일을 pandas DataFrame으로 로드 후 필요에 맞게 편집하는 방식을 권장합니다.

1️⃣ 행 단위(Line-level) 데이터 불러오기 예시

 
  from datasets import load_dataset 
  df_line = load_dataset(
  "csv", 
  data_files={ "train": "hf://datasets/AKS-DHLAB/KPoEM/KPoEM_line_dataset_v4.tsv"}, 
  delimiter="\t", 
  encoding="utf-8", 
  quoting=3, 
  ) 
  df = df_line["train"].to_pandas() 
  df.head()

2️⃣ 작품 단위(Poem-level) 데이터 불러오기 예시

 
  from datasets import load_dataset
  df_poem = load_dataset(
    "csv",
    data_files={
        "train": "hf://datasets/AKS-DHLAB/KPoEM/KPoEM_poem_dataset_v4.tsv"},
    delimiter="\t",
    encoding="utf-8",      
    quoting=3,             
    )
  df = df_poem["train"].to_pandas()
  df.head()

Contributors

• IRO LIM — Producer, Dataset Lead
• Haein Ji — Data Curator
• Sul Koo — Annotator
• Song-yi Jung — Annotator
• Jonghoon Yun — Annotator
• Byungjun Kim — Supervisor

---

• Please refer to the following presentation: LIM, I., Ji, H., & Kim, B. (2025). 한국 근현대시 감정 라벨링 데이터셋 구축: 문학 텍스트의 컴퓨터 기반 감정 분류와 생성형 AI 활용을 위한 기초 연구. 제2회 한국현대문학자대회 (KorLitConf), Seoul, Korea. Zenodo. https://doi.org/10.5281/zenodo.15055795
• All source code used in this study has been made publicly available in the following repository. See the link for details. https://github.com/AKS-DHLAB/KPoEM
• train, validation, test dataset 구분 없이 문장(line) 단위, 작품(poem) 단위 tsv 파일로 업로드하였습니다.
• Update Records: The KPoEM dataset has been updated and is permanently archived on Zenodo. DOI: KPoEM dataset v4.0 in zenodo