Barak Krakauer is a data scientist at YouTube and Google. He currently works on problems related to mitigating abuse in comments on YouTube. He has previously worked in data science, primarily in natural language processing issues, at Levi Strauss & Co and at IBM. Before that, he taught philosophy at the University of California, Santa Cruz and worked as an analyst in educational policy. He has a PhD in Philosophy from the University of Massachusetts, Amherst.

In his contribution to the forum, he’ll be talking about “Improving Natural Language Classification at YouTube”. Here’s an abstract:

There are numerous difficult problems that face any effort to keep abusive content off of YouTube, and in ensuring that the efforts to do so are implemented in a fair way. The space is full of motivated, adversarial actors who adapt to policy and enforcement changes, and many of the forms of abuse are difficult to automatically detect and evaluate. This talk is an overview of some of these issues, discussing the types of abuse seen and the challenges faced by data scientists who attempt to mitigate these harms, with a particular focus on issues related to text comments and fairness.