HateXplain: A Benchmark Dataset for Explainable Hate Speech Detection

By Binny Mathew, Punyajoy Saha, Seid Muhie Yimam, Chris Biemann, Pawan Goyal, Animesh Mukherjee. 2020

821 citations recorded by the source corpus.

Hate speech is a challenging issue plaguing the online social media. While better models for hate speech detection are continuously being developed, there is little research on the bias and interpretability aspects of hate speech. In this paper, we introduce HateXplain, the first benchmark hate speech dataset covering multiple aspects of the issue. Each post in our dataset is annotated from three different perspectives: the basic, commonly used 3-class classification (i.e., hate, offensive or normal), the target community (i.e., the community that has been the victim of hate speech/offensive speech in the post), and the rationales, i.e., the portions of the post on which their labelling decision (as hate, offensive or normal) is based. We utilize existing state-of-the-art models and observe that even models that perform very well in classification do not score high on explainability metrics like model plausibility and faithfulness. We also observe that models, which utilize the human rationales for training, perform better in reducing unintended bias towards target communities. We have made our code and dataset public for other researchers.

Subfields: Artificial Intelligence, Natural Language Processing, Safety, Robustness & Interpretability

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

In Safety, Robustness & Interpretability

Citations in the corpus

Towards Safer Generative Language Models: A Survey on Safety Risks, Evaluations, and Improvements (2023, background): “Many studies focus on the interpretability of model outputs, such as visualizing decision-making processes using attention scores (Mathew et al., 2021) or improving moral judgment accuracy using knowledge-based reasoning (Mehrabi et al., 2022; Kim et al., 2022).”
Learning from the Worst: Dynamically Generated Datasets to Improve Online Hate Detection (2021, background): “Subtle and implicit forms of hate speech can also create false negatives (Vidgen and Yasseri, 2019; Palmer et al., 2020; Mathew et al., 2020), as well as more ‘complex’ forms of speech such as sarcasm, irony, adjective nominalization and rhetorical questions (Caselli et al., 2020; Vidgen et…”
Trustworthy AI: From Principles to Practices (2021, background): “Development Functional Testing [164, 235, 381] Performance Benchmarking [40, 56, 88, 93, 123, 127, 153, 238, 248, 307, 327]”
Explainable Artificial Intelligence Applications in Cyber Security: State-of-the-Art in Research (2022, neutral): “[169] NB and RF 2020 √ √ Self explainable [171] RNN and CNN 2021 √ √ √ √ √ LIME”
Supervising Model Attention with Human Explanations for Robust Natural Language Inference (2021, methodology): “Attention-based methods to incorporate explanations have also been found to improve performance on hate speech detection (Mathew et al. 2021).”
From Pretraining Data to Language Models to Downstream Tasks: Tracking the Trails of Political Biases Leading to Unfair NLP Models (2023, background): “While other tasks and datasets (Emelin et al., 2021; Mathew et al., 2021) are also possible choices, we leave them for future work.”
SOLID: A Large-Scale Semi-Supervised Dataset for Offensive Language Identification (2020, background): “Hate speech identiﬁcation is by far the most studied abusive language detection task (Ousidhoum et al., 2019; Chung et al., 2019; Mathew et al., 2021).”
Human-Machine Collaboration Approaches to Build a Dialogue Dataset for Hate Speech Countering (2022, methodology): “Many benchmarks for automatic HS detection are currently available (Mathew et al., 2021; Cao et al., 2020; Kumar et al., 2018; Hos-seinmardi et al., 2015; Waseem, 2016; Burnap and Williams, 2016).”
ToxCCIn: Toxic Content Classification with Interpretability (2021, background): “Mathew et al. (2020) also provided a hate speech dataset where samples are annotated with rationales and other labels; though they also provided baseline models that can incorporate rationales, these models were built to evaluate the effectiveness of their proposed dataset, and thus cannot be…”
Your fairness may vary: Pretrained language model fairness in toxic text classification (2021, background): “For the HateXplain dataset, the gap between balanced accuracy and fairness variability is more prominent.”
Probing Toxic Content in Large Pre-Trained Language Models (2021, background): “Consequently, there has been an increasing amount of work on explainability for toxic language classiﬁers (Aluru et al., 2020; Mathew et al., 2021).”
Teach Me to Explain: A Review of Datasets for Explainable Natural Language Processing (2021, background): “[67] webpage relevance ranking 2-3 sentences crowd 700 (15) SCAT [139] document-level machine translation none experts ∼14K ECTHR [24] alleged legal violation prediction paragraphs auto + expert ∼11K HUMMINGBIRD [48] style classification words crowd 500 HATEXPLAIN [79] hate-speech classification phr”
On the Intrinsic and Extrinsic Fairness Evaluation Metrics for Contextualized Language Representations (2022, neutral): “Extrinsic: HateXPlain (H ATE X) (Mathew et al., 2020) measures bias in hate speech detection systems.”
Explainability for Large Language Models: A Survey (2023, background): “Understanding Understanding the interplay between training data characteristics and the resulting behavior of the model will provide key insights into the source of emergent abilities in LLMs.”