robertuito-hate-speech
| Property | Value |
|---|---|
| Base Model | RoBERTuito |
| Task | Hate Speech Detection |
| Language | Spanish |
| Performance | 75.9% Macro F1 Score |
| Source | SemEval 2019 Task 5 |
| Repository | GitHub |
What is robertuito-hate-speech?
robertuito-hate-speech is a specialized Spanish language model designed for detecting hate speech in social media content, particularly Twitter. Built upon RoBERTuito, a RoBERTa model pre-trained on Spanish tweets, this model was specifically fine-tuned using the SemEval 2019 Task 5 HatEval corpus. It excels in multi-classification tasks related to hate speech detection.
Implementation Details
The model implements a three-way classification system that analyzes text for hate speech characteristics. It leverages the robust architecture of RoBERTuito, which was trained on over 500 million Spanish tweets, ensuring strong performance on social media content.
- Built on RoBERTuito architecture, optimized for Spanish social media text
- Trained using SemEval 2019 Task 5 dataset
- Achieves state-of-the-art performance with 75.9% Macro F1 score
- Outperforms other Spanish language models in hate speech detection
Core Capabilities
- HS (Hate Speech): Determines if content contains hate speech
- TR (Target): Identifies if the speech targets specific individuals
- AG (Aggression): Analyzes the aggressive nature of the content
- Cross-lingual capabilities for English-Spanish code-switching scenarios
Frequently Asked Questions
Q: What makes this model unique?
This model combines the power of RoBERTuito's Spanish social media understanding with specialized hate speech detection capabilities. Its three-dimensional analysis (HS, TR, AG) provides a comprehensive assessment of problematic content, making it particularly valuable for content moderation.
Q: What are the recommended use cases?
The model is ideal for social media content moderation, online harassment detection, and research on hate speech in Spanish-language social media. It's particularly effective for platforms dealing with user-generated content and organizations monitoring online discourse.
