OffensiveLang: A Community Based Implicit Offensive Language Dataset

TL;DR

OffensiveLang tackles implicit offensive language detection by using prompt-driven generation with ChatGPT 3.5 to create a community-based dataset spanning 38 target groups and 7 categories. The work compares human and AI (ChatGPT) annotations, annotates via MTurk, and benchmarks multiple transformer models (notably BERT) to detect implicit hate speech, highlighting the trade-offs between precision and recall. A key contribution is the prompt-based zero-shot approach for data generation and annotation, enabling scalable data creation while addressing ethical constraints. The dataset and findings offer a practical resource for advancing implicit offensive language detection and inform future work on expanding categories and improving cross-domain generalization, with careful attention to ethics and transparency.

Abstract

The widespread presence of hateful languages on social media has resulted in adverse effects on societal well-being. As a result, addressing this issue with high priority has become very important. Hate speech or offensive languages exist in both explicit and implicit forms, with the latter being more challenging to detect. Current research in this domain encounters several challenges. Firstly, the existing datasets primarily rely on the collection of texts containing explicit offensive keywords, making it challenging to capture implicitly offensive contents that are devoid of these keywords. Secondly, common methodologies tend to focus solely on textual analysis, neglecting the valuable insights that community information can provide. In this research paper, we introduce a novel dataset OffensiveLang, a community based implicit offensive language dataset generated by ChatGPT 3.5 containing data for 38 different target groups. Despite limitations in generating offensive texts using ChatGPT due to ethical constraints, we present a prompt-based approach that effectively generates implicit offensive languages. To ensure data quality, we evaluate the dataset with human. Additionally, we employ a prompt-based zero-shot method with ChatGPT and compare the detection results between human annotation and ChatGPT annotation. We utilize existing state-of-the-art models to see how effective they are in detecting such languages. The dataset is available here: https://github.com/AmitDasRup123/OffensiveLang

Figures (7)

• Figure 1: Workflow diagram of our work. It shows the way OffensiveLang was generated texts using ChatGPT 3.5 and then was annotated with both chatGPT and human.
• Figure 2: Distribution of the seven categories in OffensiveLang.
• Figure 3: Visualization of the performances of different models. It can be seen that the BERT model achieved the highest F1 score and Recall values whereas DistilBERT achieved the highest Precision score.
• Figure 4: Example where ChatGPT prohibits generating offensive texts.
• Figure 5: Our prompt for generating community basesd implicit offensive text using ChatGPT.