Researchers develop algorithm to identify mysogynistic tweets

Twitter and other popular social platforms have a major hate speech problem. Despite Twitter’s efforts, it’s still very slow and ineffective in removing hate speech from its platform. There could be some respite as researchers have now developed an algorithm that detects misogynistic content on Twitter.

This algorithm has been developed by researchers from the Queensland University of Technology (QUT) in Australia. The process required mining 1 million tweets and then filtered these tweets using three abusive keywords - slut, rape and whore. This was done as it would be a quicker process then to read each tweet individually. The researchers also had to classify if such tweets are a joke, said in a sarcastic way or actually targeting someone or a group.

The remaining 5,000 tweets were then categorised as either misogynistic or not. A machine learning classifier was then trained with the final classified misogynistic tweets to create its own classification model to identify such tweets. The classification of misogynistic tweets aren’t limited to these keywords though. One of the co-authors Professor Richi Naya explained how a phrase like “get back to the kitchen” will be classified as not misogynistic since it doesn’t have any abusive words. The machine learning classifier has been refined to actually differentiate and understand such tweets.

ALSO READ: Social media platforms face a reckoning over hate speech

Based on the testing conducted by the researchers, the algorithm can identify misogynistic tweets with 75% accuracy. The researchers said this algorithm can be used to identify racism, homophobia and even abuse of disabled people. The researchers also hope that social media companies use the tool on their platforms.