ftan-2.5

ftan-2.5 is a fine-tuned DistilBERT sequence classification model for detecting offensive text.

The model predicts one of two labels:

• clean — non-offensive text
• offensive — offensive text

Training

ftan-2.5 is the continuation of the akaruineko/ftan-2.0 project.

The new version was trained on the akaruineko/fantastic-offensive dataset, containing approximately 2.46 million text samples with a near-balanced distribution between the clean and offensive classes.

Usage

from transformers import pipeline

classifier = pipeline(
    "text-classification",
    model="akaruineko/ftan-2.5"
)

result = classifier("you are stupid")
print(result)

Example:

[{'label': 'offensive', 'score': 0.997348427772522}]

Intended Use

ftan-2.5 can be used for:

• content moderation
• filtering offensive messages
• dataset preprocessing
• moderation pipelines

Limitations

This model should not be treated as a perfect moderation system.

Offensiveness can depend heavily on context, intent, quotation, sarcasm, reclaimed language, and the surrounding conversation. The model may therefore produce incorrect predictions for ambiguous or context-dependent text.

For example, a sentence discussing an offensive word may still receive a non-trivial offensive score even when the sentence itself is not an insult.

The model also operates on individual text inputs and does not have access to conversation history unless it is explicitly provided as input.

License

MIT

Author

Created by akaruineko.

This model is the 2.5 continuation of the ftan model series.