Human-Centered Language & Information Technologies (HUMANE) Lab at Soongsil University studies NLP and multimodal technologies based on large language models. Our research aims to tackle social problems in online platforms (e.g., misinformation, hate speech, bias) by developing learning and inference methods centered around human values.
숭실대학교 인간-중심 언어 및 정보 기술 (휴메인) 랩은 대규모 언어 모델 기반의 자연어처리 및 멀티모달 인공지능 기술을 연구합니다. 사람 가치 중심의 학습 및 추론 방법을 개발하여 온라인 플랫폼에서 나타나는 사람과 사회에 관련된 문제(예. 오정보, 혐오, 편향)를 해결하는 것을 목표합니다.
Kunwoo Park / 박 건우 (kunwoo.park at ssu dot ac dot kr)
연구 협업, 산학 협력 등 다양한 형태의 협업을 환영합니다.
Prof. Park won the Outstanding Reviewer Award at EMNLP 2024.
Sep 30th, 2024We placed runner-up at the AVeriTeC challenge, aiming to verify real-world claims using LLMs (collaboration with maum.ai and Adobe Research). Our system is invited as an oral presentation at the FEVER workshop co-located with EMNLP 2024.
May 16th, 2024A joint work with Adobe Research has been accepted to ACL 2024.
Mar 8th, 2024A paper is accepted for publication in IEEE Access. We introduced QuoteCSE++, a data-centric contrastive embedding framework for contextomy detection.
Feb 20th, 2024A collaborative work with IU has been accepted to LREC-COLING 2024. We investigated how ChatGPT is aligned with human's views on explanation quality.
Dec 21th, 2023A study on the LLM's ability of mathematical reasoning won the Outstanding Paper award at KSC 2023 (정보과학회 동계학술대회).
Oct 17th, 2023A paper is accepted to EMNLP 2023 (Findings) based on collaboration with TUNiB. We introduce K-HATERS, the largest hate speech detection corpus in Korean.
July 17th, 2023Our collaborative study with Adobe Research won the Hyundai Motors Best Paper Award at CKAIA (한국인공지능학회).
July 6th, 2023A joint work with UCLA is accepted for publication in EPJ Data Science (Title: Perceived Gender Cues from Social Media Photographs and its Relations to Electoral Success).