The DH Cookbook in Korean Studies

A Fusion of Old and New

Aron van de Pol

Leiden University

2024-10-18

Welcome to the DH Kitchen

• The Kitchen
  • The Tools: Methods and Technologies
  • The Ingredients: Data and Sources
  • The Recipes: Research Designs and Approaches

Figure 1: The DH Cookbook generated by Dall-e Model.

Goal today: Explore Digital Humanities approaches to Korean Studies.

• The Kitchen (Research Environment):
  • Traditional spaces: libraries, archives, field sites
  • Digital spaces: online databases, virtual collaboratories, computational clusters
  • Hybrid spaces: digitally-enhanced physical archives, VR/AR research environments
• The Tools (Methods and Technologies):
  • Traditional: close reading, archival research, interviews
  • Digital: text mining, network analysis, GIS mapping, data visualization
  • Computational: machine learning, natural language processing
• The Ingredients (Data and Sources):
  • Traditional: physical manuscripts, artifacts, recorded oral histories
  • Digitized: scanned documents, 3D models of artifacts, digital audio/video
  • Born-digital: social media data, digital art, online communities
• The Recipes (Research Designs and Approaches):
  • Combining traditional and digital methods to address research questions
  • Examples: distant reading of large corpora, combining quantitative analysis with qualitative interpretation, using AI to assist in translation and analysis

Our goal is to explore how these elements come together in Digital Humanities approaches to Korean Studies, showcasing innovative “recipes” that blend traditional scholarship with cutting-edge digital methods.

Recipes: What have researchers been cooking?

• Diverse flavours!
• Text Mining
• Computer Vision
• Network Analysis
• QGIS (Mapping)
• Survey/Numerical Data

Detection of Four panel cartoons in the Chosŏn Ilbo 1920-1940 Lee, Kim, and Jun¹

1. “Automatic Detection of Four-Panel Cartoon in Large-Scale Korean Digitized Newspapers Using Deep Learning”.

Network of Medieval Korean personalities Cha¹

1. “Javier Cha”.

The study of socialist tone in Kaebyŏk Hur¹

1. “『개벽』 논조의 사회주의화에 관한 새로운 접근 토픽 연결망 분석을 중심으로 A New Approach to Socialization in Gaebyeok’s Tone: Focusing on Topic Network Analysis”.

Why do anti-democratic laws exist in democracies? Green and Denney¹

1. “Why Do Democratic Societies Tolerate Undemocratic Laws?”

Multi-instance learning to extract visual identities of Colonial Print shops

LLM Driven metadata Annotation of Colonial Korean Advertisements

{
  "product": "Shoes",
  "company": "朴德裕洋靴店",
  "location": "京城府宽𤏩洞五一番地",
  "language": "Korean (mixed)",
  "visual_elements": {
    "illustration": "Man wearing shoes, 
     pointing at product illustration",
    "text": "Advertisement text in Korean"
  }
}

And much more…

• Please check out the papers of the researchers mentioned!
• Korean Journal of Digital Humanities
• See Cha & Wall (2023) for more depth/examples.¹

1. The rest of this journal volume has more in-depth papers giving a wider overview. Cha and Wall, “Introduction to Special Section Digital Korean Studies”.

The Pantry: Ingredients of DH/Korean Studies Fusion

Ingredients: Data

• The data remains mostly the same!
• Digitized/Machine Readable.
• Format changes.

Shopping for data

• Starts mostly digitally
  • Archives
  • Libraries
  • Google Scholar
  • ChatGPT/LLMs?

Shopping for data

(a) Korean History Database

(b) Document Specific Databases

(c) Hyundam Mungo Dataset

(d) Seoul Public Datasquare

Figure 2

Shopping for data

• Getting data is hard.
• Especially historical data depends on digitization and OCR¹
• Current digitization at the National Library of Korea: 19% (2023)²
• This is still 1.9 million works…
• The challenge: How to curate the data we need?
• Logistical Challenges

If we have such data available, how do we know - Which parts to use? - How large should our sample/representation be? Etc.

1. Optical Character Recognition. Moving from image to machine-readable text.

2. Kim, “Korean Memory Project”.

One bite at a time?

• Libraries/Archives have more available than ever.
• But also harder to get in bulk
• (Computational) DH requires scale.
• Most facilities are not ready to support this.

One bite at a time?

• API interfaces (For Archives/Libraries)¹
• Compute facilities for Researchers (For Research Institutes)
• How do we share the data used again?
  • Copyright
  • Long-term storage solution

1. API - Application Programming Interface allows interaction with the data behind the website, enabling researchers to programmatically access and retrieve data at scale without overloading the servers.

Naver’s Kak Data Center

• Sejong & Ch’unch’ŏn
• One ‘Sejong’
  • Can house 600,000 server units
  • 65 exabyte of data¹
  • ~1 million times the NLK²

• One ‘Ch’unch’ŏn’
  • Four Hydroelectric plants and backup generators

Naver’s Kak Data Centers. Naver³

1. 65.000.000 TB

2. Park, “[르포] 국립중앙도서관 1만개 분량 데이터 담은 네이버 ‘각 춘천’ 가보니”; Park, “로봇 가로\(\cdot\)세로가 데이터 나른다, 국립중앙도서관 ‘100만배 용량’”.

3. “DATA CENTER GAK”.

NLK Data Preservation Center P’yongCh’ang

• NLK also developing new Data Center
• Problem of Power/Electricity
• Humanist need to consider these logistical challenges now.

Computer Render of the Envisioned Data Center

Outside Print

• I work on Colonial Korea > ‘Centralized’ data.
• Increasing interest in non-print data: online communities, internet archives, social media.
• Interest in preservation and analysis of non-print data.
  • Where is this data? Where is it stored? And by Whom?

Kitchen Tools: Technologies and Platforms

Kitchen Tools: Technologies and Platforms

Software

• Python/R
• LaTeX/Typst/Quarto
• Open Refine
• Etc…

Hardware

• Computing Facility
• Long term Data storage
• VR/AR
• Research ‘Lab’

Large Language models

• Perhaps the most hyped
• Basic understanding of the mechanism
• How to use it responsibility

Choosing your cuisine

Should Everyone Cook This Way?

• Not necessarily.
• However, understanding the basics of ‘preparing a meal’ is essential.
• Mastery of core techniques important for a future with an increasing digital side.

Should Everyone Cook This Way?

• Not necessary to engage quantitatively.

• However, engaging with digital media, data, and tools is crucial.

Should Everyone Cook This Way?

• How do you browse through an archive containing
  • millions of books?
  • 100,000 years of video material?
  • Gigabytes (if not Tera) of image data?
  • CSV, plain text, of millions of rows of social media data?

Fusion Cuisine: Digital Methods, New Questions

• Scale vs. Detail.
• Traditional research & Computational Approaches look at similar questions
• Inherently digital: Some studies, are inherently digital, but not inherently quantitative
• Added value in a different way to read/view distantly.

Fusion Cuisine: Digital Methods, New Questions

• Examples:
  • Historical text analysis at scale.
  • Browsing & analysing 10,000s of image data.
  • Digital anthropology of K-pop fan communities.
  • Study of public data.

• Expanding reach: Digital methods allow us to address traditional research questions on a larger scale.
• Trade-off: We exchange fine-grained detail for scale and breadth.
• Inherently digital research: Some questions can only be explored through digital methods.
• Examples:
  • Historical Text Analysis: Digital tools reveal large-scale trends across archives, but may miss subtle nuances found in close reading.
  • Online Communities: Digital anthropology of K-pop fan communities, analyzing millions of posts, interactions, and evolving dynamics.

The Potluck: Collaborative and Open Scholarship

• Sharing “recipes”
  • Open data, methods, and code.
• Collaboration
  • Computational and non-computational working together.
• Plurality
  • Not qualitative vs. quantitative but building upon each other.

Open Fields: New Recipes Waiting to be Created

• LLM-assisted studies
• Big data in Korean Studies
• Digital preservation efforts

Conclusion: The Ever-Evolving Menu

• DH in Korean studies: A diverse and expanding cuisine
• Different research questions call for different recipes.
• Readiness for a more digitized world.

Q&A: Taste Testing

Thank you for joining today’s “feast”! What did you think of the “menu” we explored?

Questions, comments, or feedback?

https://aronvandepol.com/

An Sŏng Jae judging critically the food served in hŭkpaek yorisa. (Source: Netflix)

Cited Works

Cha, Javier. “Javier Cha.” Javier Cha. https://javiercha.com/, 2024.

Cha, Javier, and Barbara Wall. “Introduction to Special Section Digital Korean Studies.” Korean Studies 47, no. 1 (2023): 1–7.

Green, Christopher, and Steven Denney. “Why Do Democratic Societies Tolerate Undemocratic Laws? Sorting Public Support for the National Security Act in South Korea.” Democratization 31, no. 1 (January 2024): 113–31. https://doi.org/10.1080/13510347.2023.2258082.

Hur, Soo. “『개벽』 논조의 사회주의화에 관한 새로운 접근 토픽 연결망 분석을 중심으로 A New Approach to Socialization in Gaebyeok’s Tone: Focusing on Topic Network Analysis.” 인문논총 78, no. 1 (2021): 221–62.

Kim, Sujeong. “Korean Memory Project: Digital Curation of Knowledge Information Resources.” Conference {{Presentation}}. Seoul, October 2023.

Lee, Seojoon, Byungjun Kim, and Bong Gwan Jun. “Automatic Detection of Four-Panel Cartoon in Large-Scale Korean Digitized Newspapers Using Deep Learning.” Journal of Open Humanities Data 10, no. 1 (June 2024). https://doi.org/10.5334/johd.205.

Naver. “DATA CENTER GAK.” {Company Website}. DATA CENTER GAK. https://datacenter.navercorp.com/, 2024.

Park, Jiyeong. “로봇 가로\(\cdot\)세로가 데이터 나른다, 국립중앙도서관 ‘100만배 용량’.” 한겨레, November 2023.

Park, Suhyeon. “[르포] 국립중앙도서관 1만개 분량 데이터 담은 네이버 ‘각 춘천’ 가보니.” 조선비즈, February 2023.