Schema

We developed the annotation schema to maximize the annotator's efficiency.

OLID-BR contains a collection of annotated sentences in Brazilian Portuguese using an annotation model that encompasses the following levels:

• Offensive content detection
• Offense target identification
• Offensive spans identification

Hierarchical taxonomy for categorizing offensive language, proposed by author.

To achieve this, we defined 4 questions that our qualified annotators will answer to each sentence.

• Is this text toxic?
• Which kind of toxicity it has?
• There's a specific target?
• Which words make this text toxic/offensive?

The following image shows the annotation screen that our annotators will see.

Labeling Interface - Label Studio

---

Last update: March 1, 2023