Chaima-KHENAFIF/fake-news-spreader-detector

Fake News Spreader Detector

A fine-tuned bert-base-uncased model that classifies a Twitter user as a fake news spreader or not, based on their aggregated post history.

Trained on a PAN-format author profiling dataset. TF-IDF + classical models (Logistic Regression, SVM, Random Forest, Naive Bayes) were compared as baselines before fine-tuning BERT as the final approach.

Usage

from transformers import pipeline

classifier = pipeline("text-classification", model="Chaima-KHENAFIF/fake-news-spreader-detector")
classifier("aggregated user post text goes here")

Training

Fine-tuned for up to 5 epochs with early stopping (patience 1, selected on F1), on text cleaned via lemmatization, stopword/punctuation/URL/mention removal.

Results

Metric	Score
Accuracy	TODO
F1	TODO

Links

• Full code, training pipeline, and Flask interface: GitHub repo