Twitter Toxicity

web app 2019

As an individual university assignment I created a web application for analysing user’s Twitter timeline using machine learning, Python and JavaScript.

As we spend more time online, quality of that time matters a lot. What we hear, read and listenon the social networks influence our opinions.

With assumption that none of us wants to surround herself with insults and aggressive opinions, this tool was created to help analyze tweets of user’s friends and tries to measure toxicity of their language. That means that it takes tweets of people user follow on Twitter and Perspective API determines toxicity of their content.

Note that the data model is far from perfect and this project is just experiment. It works better with tweets in english and would need far more testing to provide useful results for different scenarios.

You can see how does it work in this video:

 

 

The app is written in JavaScript and Django framework. The toxicity measurement is done through Google Cloud services and the source code is available on GitHub.