Title: Three Modeling Vignettes from Search Ads Quality at Google

Speaker: Sugato Basu

Affiliation: Google

Date and location: Friday, March 8, 12:30 – 1:30 pm; Wozniak Lounge (430 Soda Hall)

Abstract: Building and deploying machine-learning (ML) models at Google comes with interesting challenges. For example, some models have to handle massive amounts of training data, while some supervised tasks have insufficient amount of training labels. Or, even when the model quality is good enough for a product requirement, it may not meet other requirements (e.g., serving latency, memory footprint). In this talk we will discuss some of these challenges and share our experiences from deploying ML models for quality improvements in Search Ads products via some case studies.

The first case study will discuss transfer learning in the context of Search Ads Quality models: can we learn a rich feature representation in a model with lots of training data, and use that in other parts of the prediction stack in Search Ads? The second case study will discuss advertiser response models: can we model how advertisers adjust their bids in response to feature launches in Google Ads system? Finally, the third case study will study performance attribution in ad text: can we identify ad text segments that predict high/low performance of the ad, discounting the effect of confounding text (e.g., brand names)? I will discuss in detail how we use a wide range of ML techniques to solve these problems, e.g., factorized models with deep and cross network stacks, sequence models with mirror attention architecture, and convolutional adversarial models.

Bio: Dr. Sugato Basu is currently the Tech Lead of the AdsAI team in Google, which applies state-of-the-art machine learning (ML) and natural language processing (NLP) technology to challenging problems in Search Ads at Google. He joined Google in 2007 and has worked for more than a decade on various ML problems in computational advertising related to user/advertiser modeling, auction, pricing, and other problems in the prediction stack. Before that he spent about a year at SRI International working on the CALO project, a precursor to Siri. Sugato did his PhD in Computer Science from the University of Texas at Austin, where his research on semi-supervised clustering got 2 best paper awards. He likes spending his free time with his wife (who is equally if not more more passionate about AI) and 9-year old daughter (who dreams of being an “AI scientist” when she grows up).