Grant Wang came up with the idea to develop a tool to detect offensive content in social media posts after he lost a friend to cyberbullying.
“I wanted to use my expertise in linguistics and computer science to try and combat this issue,” he said. “It’s becoming a greater and greater problem over time, with social media really having more and more hatred on it.”
The model he developed can detect offensive content in 100 languages, and captured the attention of a nationwide science talent competition.
“This is like a dream come true,” said the Williamsville East High School senior, one of 40 finalists chosen from more than 2,100 entrants in the Regeneron Science Talent Search 2024.
Each finalist will receive a $25,000 award and participate in a weeklong competition and rigorous judging process in March.
The top 10 will get awards that range from $40,000 to $250,000.
The competition started in 1942 as the Westinghouse Science Talent Search.
Past finalists have won 13 Nobel prizes and 22 MacArthur Foundation fellowships.
Grant heard about the competition when he was a freshman, the year he spent learning mostly remotely because of the Covid-19 pandemic.
Michael Kubiak, his AP environmental science teacher, believes awards are in store, starting with more honors in the competition next month.
“He’s one of the hardest-working students I’ve ever had,” Kubiak said. “I would not be surprised if he wins it.”
Grant’s model is different from some others because his considers the context of the post, not just an individual word, and is available in many more languages than English.
“My model does look at the context of the sentence,” he told The Buffalo News. “It understands the overall meaning of the sentence. Essentially it can understand the nuances of the sentence because the training data is robust.”
Grant started the project in mid-2022, and developed a solid model that he presented at the International…
Read the full article here
Leave a Reply