Research map / Label-efficient and robust learning / Recognition and long-tail
An Information Retrieval Approach to Building Datasets for Hate Speech Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use pooling instead of keyword search combined with random sampling to select tweets for annotation | Hate speech dataset construction · Twitter hate speech dataset | relative coverage: 10.40% -> 14.60% | Helped |
| Use pooling instead of keyword search combined with random sampling to select tweets for annotation | Hate speech dataset construction · Twitter hate speech dataset | prevalence: 4.96% -> 14.12% | Helped |
| Test on tweets without known hate words rather than tweets with them | BERT hate speech classifier · Our dataset test set | F1 (Hate): 47.79 -> 26.66 | Hurt |
| Test on tweets without known hate words rather than tweets with them | LSTM hate speech classifier · Our dataset test set | F1 (Hate): 35.07 -> 14.28 | Hurt |
| Test on tweets without known hate words rather than tweets with them | BiLSTM hate speech classifier · Our dataset test set | F1 (Hate): 34.78 -> 8.695 | Hurt |
| Use active learning (SAL or CAL) instead of passive random document selection (SPL) | Active-learning annotation selection · Retrospective evaluation on the pooling-selected documents | F1 | Helped |
| Select likely hateful tweets with CAL rather than uncertain tweets with SAL | Active-learning annotation selection · Retrospective evaluation on the pooling-selected documents | F1 | Helped |
| Represent tweets with BERT contextual embeddings rather than TF-IDF | Active-learning annotation selection · Retrospective evaluation on the pooling-selected documents | F1 | Helped |
| Annotate 50% of the pooling-selected documents using active learning | Active-learning annotation selection · Retrospective evaluation on the pooling-selected documents | F1: F_1 ≥ 0.9 | Helped |
| Annotate 50% of the pooling-selected documents using active learning | Active-learning annotation selection · Retrospective evaluation on the pooling-selected documents | prevalence of hateful documents found: 80% | Helped |
Closest papers by meaning
- Spanning the Spectrum of Hatred Detection: A Persian Multi-Label Hate Speech Dataset with Annotator Rationales (AAAI 2024 · similarity 0.93)
- Multi-Modal Discussion Transformer: Integrating Text, Images and Graph Transformers to Detect Hate Speech on Social Media (AAAI 2024 · similarity 0.88)
- TRACE: Textual Relevance Augmentation and Contextual Encoding for Multimodal Hate Detection (AAAI 2026 · similarity 0.87)
- A Multi-Task Benchmark for Abusive Language Detection in Low-Resource Settings (NeurIPS 2025 · similarity 0.87)
- Hateful Meme Detection through Context-Sensitive Prompting and Fine-Grained Labeling (Student Abstract) (AAAI 2025 · similarity 0.84)
- IndicCONAN: A Multilingual Dataset for Combating Hate Speech in Indian Context (AAAI 2024 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.