Michael Bendersky (Google Research)

Title: Hyperlocal Event Discovery

AbstractIn this talk, we will consider automatic discovery of hyperlocal events such as urban festivals, garage sales, farmers markets, local sporting events, etc. Hyperlocal event discovery is important for user modeling, especially in the context of mobile search and for improving personal assistant services (e.g., Google Now, Cortana, Siri). First, we present domain-agnostic, unsupervised, and highly scalable method for hyperlocal event extraction from unstructured web pages. Experiments with a large publicly available web corpus demonstrate that our method significantly increases extraction recall while maintaining high-precision standards. Second, we focus on hyperlocal event retrieval. Given an event, we propose a graph-based framework for retrieving a ranked list of related events that a user is likely to be interested in attending. We demonstrate that this framework addresses the semantic mismatch problem in a self-contained and principled manner and provides more relevant and diverse results than other standard retrieval methods. Finally, we conclude by discussing how current work can be further extended to a general framework for understanding geo-temporal entities. 

Milad Shokouhi (Microsoft Research)

Title: Implicit Feedback Signals in Query Formulation

Abstract: Query logs contain valuable information about how users interact with search engines. For instance, frequency and duration of clicks on search results have been widely used as implicit feedback for inferring search success. In this talk, we follow the footprints of users in the logs for inferring additional signals about search satisfaction. In particular, we show how user interactions during query formulation (and reformulation) can be interpreted as implicit feedback. We also demonstrate how these signals can be used to generate pseudo-labels for training auto-completion and voice recognition systems.

BioMilad Shokouhi is a Principal Applied Scientist at Microsoft. He moved to Bellevue Washington in 2015, after spending 8 years at Microsoft Research Cambridge. He’s also an honorary lecturer in School of Computing Science at the University of Glasgow. Milad’s research is mainly focused on information retrieval and web search problems. He’s interested in auto-completion, spoken queries, time-sensitive search, and personalization.

Shuang Yang (Operator)

Title: Conversational bots for e-commerce 

Abstract: I will talk about some lessons I've learned in building conversational bots specialized for e-commerce inside a modern messaging app on mobile. It is extremely challenging as it involves many of the unresolved fast-developing topics in AI (NLP, ML and IR), and the nature of the e-commerce application sets up a lot of special requirements and complexities that' re hard to achieve by conventional techniques.

Bio: Shuang Yang is the Chief Data Scientist at Operator Inc., where he leads the efforts in AI, Machine Learning and Data Science. Before that, he was a Staff Machine Learning Engineer and the Lead Scientist at Twitter, where he led the Machine Learning and Analytics Infrastructure team. Prior to Twitter, he worked on machine learning and predictive analytics at Microsoft Research and Yahoo! Labs. He earned his Ph.D from Georgia Institute of Technology in 2012. He has published actively at leading academic conferences and journals. He is the recipient of the ACM SIGIR 2011 Best Student Paper award, the UAI 2010 Best Student Paper award (nominated) and the PAKDD 2008 Best Student Paper award.