FNC1 Revisited: Two-Step Multilayer Perceptron based Stance Detection

---

This repository contains submission for MSCI641 Fake News Challenge Default Project. The trained model files an features are stored in trained.zip. Unzip this to use pretrained model and predict else make a new trained directory and run the main.py file.

Predict without training the models

1. git clone https://github.com/manavmehra96/fnc_stance_detection.git
2. cd fnc_stance_detection && unzip trained.zip
3. pip install -r requirements.txt
4. python main.py --train_feat n --train_model n

---

Predict with training the models

1. git clone https://github.com/manavmehra96/fnc_stance_detection.git
2. cd fnc_stance_detection && mkdir trained
3. pip install -r requirements.txt
4. python main.py --train_feat y --train_model y

The output directory contains the final predicted csv.

---

Usage

main.py [-h] [--train_feat (y/n)] [--train_model (y/n)]

optional arguments:
  -h, --help            show this help message and exit
  --train_feat - Train Features? (y/n)
                        
  --train_model - Train Model? (y/n)

---

The file structure of the repository is as follows -

├── main.py
├── data (dataset)
│   ├─**/*.csv
├── utils
│   ├─*build_model1.py
│   ├─*build_model12.py
│   ├─*features.py
│   ├─*read_data.py
│   ├─*predict.py
│   ├─*prediction.py
│   ├─*score.py
├── output
│   ├─*final_answer.csv
├── trained.zip(all trained models and features)

---

Main Dependencies

Keras==2.4.3        
nltk==3.5
numpy==1.19.0
pandas==1.0.3
scikit-learn==0.23.0
scipy==1.4.1
tensorflow==2.2.0

---

Disclaimer

The experiments were performed using a Tesla T4 GPU, 30GB memory and 8 core CPU

Credits

Manav Mehra ([email protected]) Rajbir Singh ([email protected])