Skip to content


Repository files navigation


This repository contains the code and data used in the paper titled 'Cross-SEAN: A cross-stitch semi-supervised neural attention model for COVID-19 fake news detection'


README is divided into 3 files.


  • - Contains sample data (2000 Fake, 2000 Genuine and 5000 Unlabelled) according to the format expected by the pipeline.
  • - Shell config file which ensures the end-to-end working.
  • - Prepares the data by extracting various inputs from the tweet objects containing json files.
  • - Prepares all the data according to the embedding vector inputs.
  • - Contains all the model component classes
  • - Runs the training epochs and evaluation steps
  • - Loads all the data and initialises training, followed by evaluation
  • - Contains all args which are parsed after running, along with default values.
  • - Responsible for the chrome extension API (to run, replace your Tweepy credentials in the file)
  • sentimentanalysis - A helper module to generate some tweet features in Uses a slightly modified version of williamscott701/sentimentanalysis to create the extended version of the data.

Compute Requirements:

  • 5GB GPU for training
  • 1GB for inference

How to run on test?

  • Unzip from Sample CTF Dataset folder OR
  • Add/Update data in data/covid19/ in json format, as fake.json, genuine.json, unlabelled.json. (sample data given in CTF are given here in json format)
  • Install the requirements pip install -r requirements.txt
  • Change the project and directory name in (optional)
  • Run the pipeline (changing the paths appropriately, if any of the above steps are done differently) sh ./


  • To run live service, follow README_Chrome_SEAN.txt
  • Separate readme are available for other aspects.


Please cite the paper if you are using the dataset or the model.


No description, website, or topics provided.






No releases published


No packages published