• Anurag Khanra - PES1UG19CS072
• Arvind Krishna - PES1UG19CS090

• Requirements

  • GCP account with resources available
  • The auth token for the GCP user named as key.json in API/
  • Python3, pip & virtualenv

• Usage

  • Clone the repository using git clone [email protected]:ArvindAROO/GCP-clickbait-detection.git
  • Create a virtualenv using source venv/bin/activate # ./venv/Scripts/activate
  • Activate the same with virtualenv venv # python -m venv venv
  • Install dependencies with pip3 install -r requirements.txt
  • Start the backend with cd API && python app.py
  • Load the plugin/ folder into any Chromium-based browsers extension using load unpacked option

• As soon as the button is clicked, the extension with a API call to the backend, with the link of the current website in included in the query
• The backend - A Flask app, fetches that websites by scraping the same, and separates the title & body which will be used for further processing

• Pretrained models

  • Tensorflow says "A pre-trained model is a saved network that was previously trained on a large dataset, typically on a large-scale task. You either use the pretrained model as is or use transfer learning to customize this model to a given task.
  • Using a pretrained Cloud Natural Language API to find category of the text available in both the title & the body
  • Comparing them to each other could give a reasonable accuracy about whether the body given is even related to the title

• AutoML Model

  • Google describes AutoML as _"AutoML enables developers with limited machine learning expertise to train high-quality models specific to their business needs."
  • The Dataset, found on kaggle had 32000 rows with titles of news articles which were classified as either clickbait or not clickbait
  • This dataset was cleaned and upload to AutoML with MultiLabel Classification and trained to a reasonable accuracy of 99.94% & a precision of 99.6% and deployed
  • Then our backend with fetch the title and send it to this endpoint, and get a response about it being a clickbait with its confidence
  • Model Statistics:
• The final result is the combination of both of these at 60:40 ratio

This dataset - https://www.kaggle.com/datasets/amananandrai/clickbait-dataset was cleaned, corrected and finally used as for the AutoML model Schema:

• Column: Headline: Contains the headline of the news article as a string
• Column: Clickbait: A boolean - like value which could be either "Clickbait" or "clickbaitnt" depending on either the headline being a clickbait or not

Eg:

A non clickbait article vs A clickbait article