Social Sentinel: Data Science + Social Media

Social Sentinel scans billions of public social media posts monthly on behalf of our clients: schools and universities. When threatening or harmful language is identified, the client is alerted to the original post for potential follow-up.

Social Sentinel's Data Science Team is responsible for many aspects of the company's product suite, but the question we receive most often is, "How do you reduce the noise from false positives?" Language is very complex and nuanced, but we were able to develop a model that substantially outperformed previous methods.

Polly Mangan and Tom Dinitz will discuss details of their false positive reduction model, from data acquisition to fitting to deployment and maintenance. Come join them for an interesting discussion about not only how to build a great model, but also utilize the results in a near-real-time product.