Saved in:
Bibliographic Details
Main Authors: Jeon, Minkyeong, Jeong, Hyemin, Kim, Yerang, Kim, Jiyoung, Cho, Jae Hyeon, Lee, Byung-Jun
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.13513
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918060224413696
author Jeon, Minkyeong
Jeong, Hyemin
Kim, Yerang
Kim, Jiyoung
Cho, Jae Hyeon
Lee, Byung-Jun
author_facet Jeon, Minkyeong
Jeong, Hyemin
Kim, Yerang
Kim, Jiyoung
Cho, Jae Hyeon
Lee, Byung-Jun
contents Language detoxification involves removing toxicity from offensive language. While a neutral-toxic paired dataset provides a straightforward approach for training detoxification models, creating such datasets presents several challenges: i) the need for human annotation to build paired data, and ii) the rapid evolution of offensive terms, rendering static datasets quickly outdated. To tackle these challenges, we introduce an automated paired data generation pipeline, called K/DA. This pipeline is designed to generate offensive language with implicit offensiveness and trend-aligned slang, making the resulting dataset suitable for detoxification model training. We demonstrate that the dataset generated by K/DA exhibits high pair consistency and greater implicit offensiveness compared to existing Korean datasets, and also demonstrates applicability to other languages. Furthermore, it enables effective training of a high-performing detoxification model with simple instruction fine-tuning.
format Preprint
id arxiv_https___arxiv_org_abs_2506_13513
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle K/DA: Automated Data Generation Pipeline for Detoxifying Implicitly Offensive Language in Korean
Jeon, Minkyeong
Jeong, Hyemin
Kim, Yerang
Kim, Jiyoung
Cho, Jae Hyeon
Lee, Byung-Jun
Computation and Language
Language detoxification involves removing toxicity from offensive language. While a neutral-toxic paired dataset provides a straightforward approach for training detoxification models, creating such datasets presents several challenges: i) the need for human annotation to build paired data, and ii) the rapid evolution of offensive terms, rendering static datasets quickly outdated. To tackle these challenges, we introduce an automated paired data generation pipeline, called K/DA. This pipeline is designed to generate offensive language with implicit offensiveness and trend-aligned slang, making the resulting dataset suitable for detoxification model training. We demonstrate that the dataset generated by K/DA exhibits high pair consistency and greater implicit offensiveness compared to existing Korean datasets, and also demonstrates applicability to other languages. Furthermore, it enables effective training of a high-performing detoxification model with simple instruction fine-tuning.
title K/DA: Automated Data Generation Pipeline for Detoxifying Implicitly Offensive Language in Korean
topic Computation and Language
url https://arxiv.org/abs/2506.13513