SVM clickbait classifier

Code and Dataset used in the paper titled, Stop Clickbait: Detecting and Preventing Clickbaits in Online News Media at 2016 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM)

If you are using this code or dataset for any research publication, or for preparing a technical report, you must cite the following paper as the source of the code and dataset.

Abhijnan Chakraborty, Bhargavi Paranjape, Sourya Kakarla, and Niloy Ganguly. "Stop Clickbait: Detecting and Preventing Clickbaits in Online News Media”. In Proceedings of the 2016 IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining (ASONAM), San Fransisco, US, August 2016.

BibTex:

@inproceedings{chakraborty2016stop,
title={Stop Clickbait: Detecting and preventing clickbaits in online news media},
author={Chakraborty, Abhijnan and Paranjape, Bhargavi and Kakarla, Sourya and Ganguly, Niloy},
booktitle={Advances in Social Networks Analysis and Mining (ASONAM), 2016 IEEE/ACM International Conference on},
pages={9--16},
year={2016},
organization={IEEE}
}

Requirements

1. JDK 1.7 or greater
2. Python modules

• numpy
• scipy
• SocketServer
• Scikit Learn
• networkx

Usage

• Download Stanford CoreNLP suite (ensure Java version compatibility) and extract
• Download python module stanford_corenlp_pywrapper
• Install python module stanford_corenlp_pywrapper following instructions in thier README.md
• In file stanford_server.py, change path to the Stanford CoreNLP suite to where the suite was extracted
• Run Command : python stanford_server.py
• On a separate Terminal, run command: python clickbait_classifier.py
• At the prompt, enter the title to be classified, or enter q/Q to exit

Code

• dataset: This directory contains both clickbait and non-clickbait headlines used to train the classifier
• dependencies : Includes the corpus of hyperbolic words, common ngrams etc.
• vectors: Includes pretrained vectors used for classification
• experiment.py: code used to run certain experiments for the paper (can be ignored)
• stanford_server.py : Exposes Stanford CoreNLP as a service on localhost
• clickbait_classifier.py : The clickbait classifier
• utility.py : Helper functions