Improving the health of public conversations on Twitter with TensorFlow
Who is this presentation for?
- Machine learning engineers, security and privacy engineers, data scientists, and TensorFlow users
“We’re committing Twitter to help increase the collective health, openness, and civility of public conversation, and to hold ourselves publicly accountable towards progress.” (Jack Dorsey, Twitter CEO)
Machine learning techniques have been increasingly successful in helping Twitter detect abusive content and promote healthy conversations. Twitter currently focuses on conversational health and information integrity. Conversational health means when you converse on Twitter, the company wants to ensure that you can have respectful conversations with genuine people. Twitter policies and their enforcement provide the boundaries of the conversations it wants on its platform, and Twitter relies on machine learning and TensorFlow to improve the health of public conversations. Information integrity means that when you try to stay informed, Twitter wants to provide you with relevant and high-quality information. The company wants it to be free from spam and irrelevant information. Machine learning has also played an important role in combating fast-evolving online adversaries.
Li Xu and Yi Zhuang examine how Twitter uses TensorFlow to improve conversation health and information integrity by detecting abusive, toxic, and spammy content and promotes healthy conversations on the platform. They detail the challenges the company faces, how it formulates them as machine learning problems, models used to tackle these ML problems, and how the company uses TensorFlow to train and serve these models in production.
- Familiarity with machine learning and TensorFlow
What you'll learn
- Understand how Twitter helps to increase the collective health, openness, and civility of public conversations and the unique challenges it faces when aiming to accomplish this
- Learn how to formulate an abstract issue as a concrete machine learning optimization problem; how Twitter uses TensorFlow to define, train, and serve ML models for the above problem; and how to productionize TensorFlow at a Twitter scale
Li Xu is a software engineer on the health machine learning team at Twitter, working on the development of machine learning technologies for health, security, and privacy. Previously, he was a software engineer on the security machine learning platform team at Uber, working on the architecture development of machine learning platform for security, and a researcher at Yahoo Labs, where he conducted state-of-the-art research on security, privacy, and machine learning. Li has shipped many inventions and technologies to Yahoo, Uber, and Twitter products. Nowadays, more than a billion users are using these products. His research interests lie in security and machine learning. He’s authored or coauthored papers in top-ranked journals, conferences, book chapters, and US patents. He served as a program committee member for top conferences of security, AI, and big data.
Yi Zhuang is a senior staff machine learning software engineer at Twitter, where he leads a team building a platform for working with ML models. He works on uniting ML practitioners around a single ML platform, bringing consistency to ML practices at Twitter. Previously, Yi led a team to develop a trillion-document-scale distributed search engine at Twitter. Yi holds an MS in computer science from Carnegie Mellon University. He loves cats and enjoys pondering over all things technical and logical.
Leave a Comment or Question
Help us make this conference the best it can be for you. Have questions you'd like this speaker to address? Suggestions for issues that deserve extra attention? Feedback that you'd like to share with the speaker and other attendees?
Join the conversation here (requires login)
For conference registration information and customer service
For more information on community discounts and trade opportunities with O’Reilly conferences
For information on exhibiting or sponsoring a conference
For media/analyst press inquires