With her algorithm, Indian professor in Australia fights misogyny online | India News

Richi Nayak has developed a new algorithm that can automatically spot and report misogynistic posts on social …Read More

NEW DELHI: Richi Nayak, a computer science professor and machine learning expert, has developed a new algorithm that can automatically spot and report misogynistic posts on social media platforms.
“At present, the onus is on the user to report abuse. Our machine-learning solution can identify and report this content to protect women online,” the professor from Queensland University of Technology, Australia, told TOIon Saturday. Details of the new algorithm were published recently in the journal Springer Nature.
Nayak, a postgraduate from IIT Roorkee, was looking to probe how the technical field of machine learning — which focuses on developing systems that can use data to learn and improve their ability overtime — can help a social cause-driven problem. Research has shown that online harassment can have devastating psychological effects on women. An Amnesty International IPSOS MORI poll in 2017 said women reported stress, anxiety or panic attacks as a result of harmful online experiences.
Nayak knew if she could make identifying and removing such content easier, it could help create a safer online space for women. Nayak’s team, including research fellow Md Abul Bashar, set out to develop a machine learning algorithm which could pick up on misogynistic and abusive words.
To make their algorithm accurate, they trained it to understand context, and to some extent, intent, behind what is being said. “A machine learning algorithm relies on data that is used to train it. The main challenge in misogynistic tweet detection is understanding the context of a tweet,” said Nayak.
Nayak’s collaborators from law department developed rules to highlight tweets that were misogynistic. “We made the model learn standard language by training with datasets like Wikipedia. Next, we trained it to learn somewhat abusive language through user reviews data. Lastly, we trained the model on a large dataset of tweets. After it had developed linguistic capability, we taught it to distinguish between misogynistic and nonmisogynistic tweets.”

Source link

Leave a Reply

Your email address will not be published. Required fields are marked *