Detoxify: A framework for abusive text transformation using LLMs

2025/07/14 by R. Chandra, Chandra, Rohitash, Choi, Jiyong +1
Computer Science · #Artificial Intelligence (cs.AI) #Computation and Language (cs.CL) #FOS: Computer and information sciences #Natural Language Processing Techniques #Software Engineering Research #Web Application Security Vulnerabilities

paper · pdf · doi:10.48550/arxiv.2507.10177

openalex publication_date 2025/07/14 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Although Large Language Models (LLMs) have demonstrated significant advancements in natural language processing tasks, their effectiveness in the classification and transformation of abusive text into non-abusive versions remains an area for exploration. In this study, we present Detoxify: a framework that employs LLMs to transform abusive text (tweets and reviews) containing hate speech and profanity into non-abusive text while retaining the original intent. We evaluate the performance of four state-of-the-art LLMs, such as Gemini, GPT-4o, DeekSeek and Groq, on their ability to identify abusive text. We aim to transform and obtain a text that is clean of abusive and inappropriate content, but maintains a similar level of sentiment and semantics, i.e. the transformed text needs to maintain its message. Afterwards, we evaluate the raw and transformed datasets with sentiment analysis and semantic analysis. Our results show Groq provides vastly different results when compared with other LLMs. We have identified similarities between GPT-4o and DeepSeek. Groq stood out as the most distinct, as it often restructured sentences with excessive positive phrasing, with the original context lost or altered.

Citations

• Chinese Toxic Language Mitigation via Sentiment Polarity Consistent Rewrites
• An evaluation of LLMs and Google Translate for translation of selected Indian languages via sentiment and semantic analyses
• The Llama 3 Herd of Models
• Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context
• DetoxLLM: A Framework for Detoxification with Explanations
• Detecting LLM-assisted writing in scientific communication: Are we there yet?
• A Comprehensive Overview of Large Language Models
• Robust Hate Speech Detection in Social Media: A Cross-Dataset Empirical Evaluation
• Don't Take This Out of Context! On the Need for Contextual Models and Evaluations for Stylistic Rewriting
• ChatGPT is not all you need. A State of the Art Review of large Generative AI models
• Detoxifying Text with MaRCo: Controllable Revision with Experts and Anti-Experts
• Estimating Social Influence from Observational Data
• Detection of Hate Speech using BERT and Hate Speech Word Embedding with Deep Model
• Text Detoxification using Large Pre-trained Neural Models
• A systematic review of Hate Speech automatic detection using Natural Language Processing
• Maximality of bi-intuitionistic propositional logic
• Learning from the Worst: Dynamically Generated Datasets to Improve\n Online Hate Detection
• HateBERT: Retraining BERT for Abusive Language Detection in English
• SemEval-2020 Task 12: Multilingual Offensive Language Identification in Social Media (OffensEval 2020)
• Language Models are Few-Shot Learners
• Med-BERT: pre-trained contextualized embeddings on large-scale structured electronic health records for disease prediction
• A review on the long short-term memory model
• MPNet: Masked and Permuted Pre-training for Language Understanding
• Re-balancing Variational Autoencoder Loss for Molecule Sequence Generation
• Tackling Online Abuse: A Survey of Automated Abuse Detection Methods
• UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction
• Large Scale Crowdsourcing and Characterization of Twitter Abusive Behavior
• An Introduction to Convolutional Neural Networks
• Adam: A Method for Stochastic Optimization
• Efficient Estimation of Word Representations in Vector Space
• Multimodal Deep Learning
• Framewise phoneme classification with bidirectional LSTM and other neural network architectures
• Long Short-Term Memory
• Users of the world, unite! The challenges and opportunities of Social Media

Related

• Efficient Models for the Detection of Hate, Abuse and Profanity
• Harnessing Artificial Intelligence to Combat Online Hate: Exploring the Challenges and Opportunities of Large Language Models in Hate Speech Detection
• LLM-based Semantic Augmentation for Harmful Content Detection
• Probing LLMs for hate speech detection: strengths and vulnerabilities
• "To Target or Not to Target": Identification and Analysis of Abusive Text Using Ensemble of Classifiers