Rude/Offensive Language Detection using Deep Learning
Detect Offensive/Rude/Toxic language from real-time tweets fetched from twitter API.
Online communication platforms and social networks like Facebook, Twitter, Instagram are overwhelmed by rude/offensive comments and useless trolling. This application fetches real-time tweets from the Twitter API and then classifies those tweets according to the offensive content in them. For classification, a pre-trained Deep Neural Net (tensorflow.js model) is used. The prediction happens in the browser itself. Once you have loaded the page/model, no other network call is made for classification.
Uses: Real time tweet monitoring. Monitoring youtube comments. Keeping out trolls and bullies and having a decent conversation on social platform(s).
Just open the application URL. It will take 20-30 seconds for the initial model to load. Kindly wait and let it load. Once you see the initial classification results, the application is good to use. DO NOT interact with the application before you see the initial results. Just fetch random tweets and see their toxic content. There is a text box at the bottom of the page where you can enter your own sentences and run the classifier on them :). Chrome on web is the best platform for checking it out.
Tensorflow.js, ReactJS, NodeJS, twit and Twitter Search API
This barely worked for me.
Even after waiting for a minute, the results were often „-”.
Since you are using Tensorflow, it would have helped to show the confidence level next to a TRUE/FALSE result.
Oh, and next time you might want to remove @-mentions from the tweets so to not ease the way of spotting the author of that tweet.
Entering manual text had no visible effect (other than blanking my input).
Good work! I just feel like it's been done before and doesn't serve any real purpose in this form. But pulling from actual tweets is a good idea and y'all executed this super well :)
This is decent work, but this is too research focused. Maybe the use case would have been something else
Nice one.. Sounds it's also better to have progress bar indicator like view. Hope you will check our entry too