Our First Podcast Features "Hope Speech" Researcher Ashique KhudaBukhsh
We finally dipped our quarantined toes into the ever-widening podcast ocean last week, because we just didn't have enough to do around here. But seriously, after more than two decades on this beat, it really seemed like the right time to start sharing some of the amazing conversations I get to have on a daily basis with the brilliant and inventive people driving high tech.
We were lucky to have as our first guest Ashique KhudaBukhsh, a project scientist in the School of Computer Science at Carnegie Mellon University's Language Technologies Institute (LTI). I met Ashique in January, when he was still a post-doctoral researcher. I stumbled upon one of his team's published papers, and I called him to talk about what they were up to. That conversation led to two stories in ADTmag's sister publication, Pure AI.
Ashique and his team are engaged in a unique and compelling line of research. He and his colleagues are using artificial intelligence (AI) to analyze online comments in social media and pick out those that defend or are sympathetic to disenfranchised groups. That research led to the development of machine learning classifiers that effectively sort the "hopeful" and "helpful" from the hateful on social media.
The LTI researchers focused initially on finding supporting content about the Rohingya people, who began fleeing Myanmar in 2017 to avoid ethnic cleansing. Ashique explains why that group was chosen, and how his team used the fastText text representation and classification library with polyglot embedding. He also explains how they developed an original strategy they call "active sampling," which used the nearest neighbors in the comment-embedding space to construct a classifier able to detect comments defending the Rohingyas among larger numbers of disparaging and neutral comments.
He also talks about how Facebook, Twitter, YouTube, and other social media platforms, which are employing strategies to identify hate speech and misinformation on their platforms, could use his team's machine learning classifiers to complement that effort.
Ashique is teaching now, but his research continues, and he came to the podcast ready to share his story. It's a great story. You should check it out.
The WatersWorks Podcast will be available soon on iTunes and other podcast apps. But you can listen to it now on the Pure AI website. While you're there, feel free to read the two stories about Ashique's team's work ("Carnegie Mellon Uses AI To Counter Hate Speech with 'Hope Speech'" and "Carnegie Mellon Continues its Research on "Hostility-Diffusing, Peace-Seeking Hope Speech"). They include links to his group's research papers, which you also might want to read.
We'll be podcasting twice a month. I'll let you know when we finish the next one.
Posted by John K. Waters on September 3, 2020 at 7:46 AM