Assistant Professor · Division of English, Sogang University

Language, people,
and the AI that works with them.

Yoonseok Heo 허윤석

I study how humans and AI collaborate: building language technology grounded in linguistic and humanistic insight, and helping students in the humanities become confident, critical users and builders of AI.

Yoonseok Heo
HAICoLab · J825

News

  • 2026.09.07ProjectETRI 용역과제 「웹 공개정보 기반 방송·미디어 콘텐츠 메타데이터 구축 자동화 및 지식자원 구조화를 위한 통합 프레임워크 개발」 착수 (2026.09 – 2026.11)
  • 2026.09.01Teaching2026학년도 2학기 개강 — AI와 언어학, 인공지능을 위한 데이터분석, 생성형 AI의 이해와 활용 강의 시작
  • 2026Paper[Paper Accepted] CulturePRM: A Process Reward Model for Mitigating Cultural Overriding in Cultural Reasoning — accepted to EMNLP 2026 (Main)
  • 2026Award[Award] 학부생 이서우·남고은 학생, KCC 2026 학부 우수논문상 (장려) 수상 — 헤밍웨이 단편 소설에 대한 LLM 기반 서사 기능 전이 패턴 분석
  • 2026.03Teaching2026학년도 1학기 신규 강의 「언어기술을 위한 데이터 중심 AI 개론」 (학부/대학원) 개설
  • 2025.09LabHAICoLab (Human-AI Collaboration Lab for Data-Centric Research) officially launched at Sogang University
About

Bridging linguistics, the humanities, and AI

서강대학교 영문학부(Division of English)에서 언어와 인공지능의 접점을 연구하고 가르칩니다. 인간-AI 협업(Human-AI Collaboration), AI 리터러시, 인간 중심 AI, 자연어처리를 중심으로 연구하며, 언어학적·인문학적 통찰이 AI 시스템의 설계와 평가를 어떻게 이끌 수 있는지에 관심이 있습니다.

Before joining the Division of English, my research focused on core NLP and multimodal learning: word sense disambiguation, Korean named entity recognition, multimodal machine translation, audio-visual scene-aware dialog, scene graph reasoning, and controllable video captioning. Recent work moves toward culturally grounded reasoning in large language models, process reward models, persona-based data augmentation, and LLM-driven analysis of literary text.

In the classroom I teach students in the humanities to read, question, and build with AI — from vibe coding and data analysis to prompt engineering and benchmark construction — so that humanistic expertise becomes a source of value in the AI era.

Research interests
Human-AI CollaborationAI LiteracyHuman-centered AINatural Language ProcessingCultural Reasoning in LLMsMultimodal LearningData-Centric AIComputational Literary Studies
At a glance
  • Position
    Assistant Professor, Department of English Literature and Linguistics, Division of English, Sogang University
  • Lab
    HAICoLab — Human-AI Collaboration Lab for Data-Centric Research (Members)
  • Office
    J825, Sogang University
  • Email
  • Phone
    02-705-8295
  • Education
    Ph.D. Computer Science and Engineering, Sogang University (2023)
    Visiting Researcher, Carnegie Mellon University (2022)
    M.S. / B.S. Computer Science and Engineering, Sogang University (2018 / 2016)
  • Profiles

Research Areas

Four threads that connect my past work in NLP and multimodal learning with current questions about how people and AI think together.

🤝

Human-AI Collaboration & Human-centered AI

How people and language models can collaborate productively — interactive, controllable systems designed around human needs rather than model capability alone.

Controllable GenerationInteractive Systems
🌏

Cultural Reasoning & Alignment of LLMs

Understanding when models recognize local cultural norms yet override them mid-reasoning, and building process reward models and persona-aware data to fix it.

Process Reward ModelsData Augmentation
🎬

Multimodal Language Understanding

Grounding language in vision and audio: multimodal machine translation, audio-visual scene-aware dialog, scene graph reasoning, and video captioning.

Vision-LanguageDialog
📚

AI Literacy & AI for the Humanities

Data-centric AI education for humanities students, and LLM-based analysis of literary text — for example narrative function transitions in Hemingway.

AI EducationComputational Literary Studies

Selected Publications

View all →
Contact

Get in touch

연구 협력, 강의 관련 문의, 학부·대학원 연구 참여에 관심 있는 학생은 이메일로 연락해 주세요. Office hours are by appointment via email.

  • Office
    J825, Sogang University
    35 Baekbeom-ro, Mapo-gu, Seoul 04107
  • Email
  • Phone
    02-705-8295
  • Office hours
    Weekdays 10:00–18:00, by appointment via email