BERT-Crash-nonCrash-Classification

Fine-tuned BERT for detecting crash-related Persian social media texts.

📄 Paper: Extracting traffic crash information from social media: an LLM-based approach – Transportation Letters (2026)

---

🎯 What it does

Binary classifier to determine whether a given Persian social media text is related to a traffic crash or not.

⚙️ Fine-tuning

• Base Model: bert-base-multilingual-cased
• Data: Proprietary Persian social media crash dataset (Damavand County, Iran)

📊 Performance

Task	Metric	Score
Crash Detection (Binary)	Accuracy	91.1%

🚀 Quick Start

from transformers import AutoModelForSequenceClassification, AutoTokenizer

model = AutoModelForSequenceClassification.from_pretrained("crash-information-extraction/BERT-Crash-nonCrash-Classification")
tokenizer = AutoTokenizer.from_pretrained("crash-information-extraction/BERT-Crash-nonCrash-Classification")

text = "تصادف در خیابان آزادی ۲ کشته داشت"
inputs = tokenizer(text, return_tensors="pt")
outputs = model(**inputs)
predicted_class = outputs.logits.argmax().item()