Multilingual Sarcasm Detector

🚀 Multilingual Sarcasm Detector

The Multilingual Sarcasm Detector is a text classification model designed to detect sarcasm in news article titles. It is fine - tuned on [bert - base - multilingual - uncased](https://huggingface.co/bert - base - multilingual - uncased). The training data includes ready - made datasets from Kaggle and scraped data from various English, Dutch, and Italian newspapers.

Labels: 0 -> Not Sarcastic; 1 -> Sarcastic

✨ Features

Source Data

• Datasets:
  • English language data: [Kaggle: News Headlines Dataset For Sarcasm Detection](https://www.kaggle.com/datasets/rmisra/news - headlines - dataset - for - sarcasm - detection).
  • Dutch non - sarcastic data: [Kaggle: Dutch News Articles](https://www.kaggle.com/datasets/maxscheijen/dutch - news - articles)
• Scraped data:
  • Dutch sarcastic news from De Speld
  • Italian non - sarcastic news from Il Giornale
  • Italian sarcastic news from Lercio

Training Dataset

• helinivan/sarcasm_headlines_multilingual

Codebase

• Git Repo: [Official repository](https://github.com/helinivan/multilingual - sarcasm - detector)

💻 Usage Examples

Basic Usage

from transformers import AutoModelForSequenceClassification
from transformers import AutoTokenizer
import string

def preprocess_data(text: str) -> str:
   return text.lower().translate(str.maketrans("", "", string.punctuation)).strip()

MODEL_PATH = "helinivan/multilingual-sarcasm-detector"

tokenizer = AutoTokenizer.from_pretrained(MODEL_PATH)
model = AutoModelForSequenceClassification.from_pretrained(MODEL_PATH)

text = "CIA Realizes It's Been Using Black Highlighters All These Years."
tokenized_text = tokenizer([preprocess_data(text)], padding=True, truncation=True, max_length=256, return_tensors="pt")
output = model(**tokenized_text)
probs = output.logits.softmax(dim=-1).tolist()[0]
confidence = max(probs)
prediction = probs.index(confidence)
results = {"is_sarcastic": prediction, "confidence": confidence}

The output of the above code is:

{'is_sarcastic': 1, 'confidence': 0.9374828934669495}

📚 Documentation

Performance

Performance Table