Hindi fake News Generation and Classification

Setting up the Repository

1. Clone the repository

git clone [email protected]:div5252/hindi-fake-news.git

2. Install the required dependencies in your python environment

pip install requirements.txt

News articles scrapping

1. Navigate to the news_scrapper directory

cd news_scrapper

2. Make a new directory data where downloaded news article will be stored

mkdir data

3. Run the navbharat_scrapper.py script

python navbharat_scrapper.py

Fake News Generation

Link to the dataset - https://drive.google.com/drive/folders/1YoEf0FxC_TNIgVlakFE3HjT0LK9EIkv2?usp=sharing

Using Split and Merge

1. Navigate to the fake_news_generator directory

cd fake_news_generator

2. Run the split_and_merge.py script

python split_and_merge.py

Using NER replacement

1. Navigate to the fake_news_generator directory

cd fake_news_generator

2. Run the .py script

python ner.py --person_list=<csv containing hindi names> --location_list=<csv containing hindi locations> --organisation_list=<csv containing hindi organization names> --input_file=<input real news csv> --output_file=<output file name> --num_steps=<number of steps at which writing takes place>

Using POS replacement

1. Navigate to the fake_news_generator directory

cd fake_news_generator

2. Run the pos.py script

python pos.py --input_file=<input real news csv file> --output_file=<output file containing the results>

Fake News Classification

Classification using similarity features

1. Navigate to the fake_news_classifier directory

cd fake_news_classifier

2. Run the news_similarity_classifier.py file

python news_similarity_classifier.py --train_path=<train dataset csv> --dev_path=<dev dataset csv> --test_path=<test dataset csv> --gold_path=<gold dataset csv> --result_path=<output text file> [--use_sentiment_features=<whether use sentiment features or not>] [--bert_dir=<bert model file for similarity features>] 

Classification using BERT

1. Navigate to the fake_news_classifier directory

cd fake_news_classifier

2. Run the bert_classifier.py file

python bert_classifier.py --train_data=<train dataset csv> --test_data=<test dataset csv> --save_dir=<output directory> --num_epochs=<number of epochs>