Hateraid

Dashboard

Inspiration

Hate speech is a well-known problem, and countering it via automatic methods can have a big impact on people’s lives Consequently, artificial intelligence plays an ever more important role in mitigating important social problems, such as the prevalence of hate speech

What it does

We created a web application that detects hate content on the internet. We analyze data from the text as well as the image to identify the true sense of the content. We also take the tweets directly from Twitter and identify if it is hate content or not.

How we built it

Once the user uploads a meme on our website, we first extract the image features along with that we are extracting textual features. These features are further passed to a deep learning-based classifier which is based on Multimodal Bitransformer which predicts it as hateful or not. At last, feedback is taken from the user to improve model performance and we retrain the model. These services are hosted on Azure cloud services.

Challenges we ran into

Technologies that are currently existing are only focusing only on one thing either textual data or image. Extracting both these features are predicting the context of the meme is difficult, for that purpose we have used Multimodal Bitransformer

Accomplishments that we're proud of

Current competition in the market is that there are start-ups that are focusing on hate detection, but they are only using either textual data or images to analyse hate But we are analysing both textual as well as image features together to make accurate predication which gives us the edge