muril-cyberbullying-detection

MuRIL Multilingual Cyberbullying Detection (v2)

This model is a fine-tuned version of Google's MuRIL (Multilingual Representations for Indian Languages) BERT architecture, specifically adapted for multi-class Cyberbullying & Hate Speech Detection.

It categorizes social media text and online commentary across 6 categories:

1. age: Cyberbullying targeting an individual's age.
2. ethnicity: Bullying or hate speech targeting ethnicity, race, or caste.
3. gender: Misogyny, sexism, or gender-based harassment.
4. religion: Hate speech or insults targeting religious beliefs.
5. other_cyberbullying: General toxic harassment, insults, or threats.
6. not_cyberbullying: Benign, safe, neutral, or positive text.

---

Model Performance

Evaluated on the held-out multi-class test benchmark:

Per-Class F1 Breakdown:

• Age: 97.76% F1
• Ethnicity: 95.86% F1
• Religion: 95.03% F1
• Gender: 86.32% F1
• Not Cyberbullying: 64.63% F1
• Other Cyberbullying: 60.11% F1

---

Quickstart & Inference

import torch
from transformers import AutoTokenizer, AutoModelForSequenceClassification

repo_id = "suyashsahu00/muril-cyberbullying-detection"

# Load model and tokenizer
tokenizer = AutoTokenizer.from_pretrained(repo_id)
model = AutoModelForSequenceClassification.from_pretrained(repo_id)

text = "Your message or tweet goes here"

# Tokenize input
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=128)

# Run prediction
with torch.no_grad():
    outputs = model(**inputs)
    probs = torch.softmax(outputs.logits, dim=-1)[0]
    predicted_idx = torch.argmax(probs).item()

predicted_label = model.config.id2label[predicted_idx]
confidence = probs[predicted_idx].item()

print(f"Prediction: {predicted_label} ({confidence * 100:.2f}%)")

---

Architectural Details

• Base Model: google/muril-base-cased
• Output Classes: 6 classes with calibrated id2label mappings
• Weights Format: SafeTensors (model.safetensors)
• Max Sequence Length: 128 tokens