NLP-Deception-Classifier

• A deception classifier using logistic regression with TF-IDF, n-grams, and sentiment analysis.
• Evaluates performance via 5-fold cross-validation and confusion matrix analysis.
• Achieved 67% accuracy on a hidden Kaggle dataset.

Usage

pip install -r requirements.txt (or conda install --file requirements.txt)
python3 train_deception_classifier.py --datafile diplomacy_cv.csv  
python3 test_deception_classifier.py --train diplomacy_cv.csv --test diplomacy_kaggle.csv --output submission.csv  

Author

Bridget Brinkman
GitHub: @F5F0A0
Project for CS 1671: Human Language Technologies, Spring 2025.
Data sourced from the Diplomacy deception corpus (Peskov et al., 2020).