Detecting and Filtering out the most annoying aspect of Youtube - Clickbaits! The tedious problem of clickbaits remains unsolved till date. bits_please proposes a user-friendly way to help out Youtube users avoid the clickbaits and unwanted youtube content with help of start-of-the-art technologies.
Solution
Detection of youtube clickbait content and highlighting the same by rendering the data for them in a different manner using a chrome extension.
Tracks
- Machine Learning - SVM Classifier
- Cloud Computing - Microsoft Azure
- Web Development
Technologies Used
- Scikit-learn
- Youtube API
- Rest API - Flask
- Web Scraping
- Client Side Rendering - Javascript
- Chrome Extension
- Pickle
Implementation :
- Collecting the Datasets : The dataset used is https://github.com/alessiovierti/youtube-clickbait-detector
- Preprocessing the Dataset:
- tokenizing
- stopwords are removed
- words shorter than 2 characters are removed
- Fitting the SVC Classifier with Dataset
- Creating a layer of flask server on top of the trained model, which will provide REST endpoints to retrieve the response.
- Flask server is deployed on Microsoft Azure.
- Creating a chrome extension using manifest.json for configuration.
- The extension will inject the content script when a certain url provided in manifest.json is triggered in the browser.
- The extension will make POST requests to https://ytclickbait.azurewebsites.net REST endpoint and will make changes to the webpage based on the returned response.
Pre requisites
- Google Chrome
- An active internet connection
Usage
- Open google chrome
- Open the extension manager by navigating to chrome://extensions from the address bar or select Extensions from More Tools from Customise and Control Google Chrome (3 dots on top right corner of the window).
- In extension manager make sure to turn on Developer Mode then click on "Load Unpacked" button and select "chrome-extension" folder downloaded from the project folder.
- Restart Google Chrome.
- Open www.youtube.com.
- Type the query in search bar and press "Enter" or click Search button.
- Extension detects and highlights the clickbait content based on trained model hosted on Microsoft Azure.
Troubleshooting
- Reload the webpage if the extension doesn't work.
- Explore console to see if extension is working fine.
Issues
- Since extension is in preview build user might have to wait for a while for the results to be displayed.
- Clicking on suggestion box to search doesn't work as of now.
- Extension works only for search queries entered in search box.
- Extension returns results for top 20 videos.