ClickbaitIndo-textclassifier

A research project to classify Indonesian online news headline into clickbait or not. Using Tensorflow Keras API and Multilingual BERT. The juPYter notebook contains all the code to train the model (and evaluate it). Labeled and unlabeled datasets are available for other researchers / enthusiasts to play with.

primary-dataset.csv (for additional evaluation) are labeled by 4 researchers and averaged to reach final category (kappa = 0.7).

Academic article is available on arXiv and currently under peer-review process at a journal.

Usage Guide

Please check API readme to use this project (courtesy of Wildan Gunawan)

Updates

update: A newly available dataset is now used for this project (credit to: William, A., & Sari, Y. (2020). CLICK-ID: A novel dataset for Indonesian clickbait headlines. Data in brief, 32, 106231.)

update: A paper regarding the design and development of the deployed model will be written soon

update 2/2/21: Wildan Gunawan contributes to deploy the model and developed the API. Kudos to him.

update: The details regarding using this model as a web app is inside the API folder (with its respective readme). You are free to use and modify the model for academic purposes only

update: The app is now accessible through our GH-Page