Title: Barbara Made the News: Mining the Behavior of Crowds for Time-Aware Learning to Rank
By: Flávio Martins (DI-FCT-UNL)
Host: Multimodal Systems
Abstract:
What is happening now propagates quickly through the Web and prompts users on the Web to interact with and produce new posts about newsworthy topics giving rise to trending topics in social-media platforms. We propose to leverage on this behavioral dynamics to estimate the most relevant time periods for a topic. Our hypothesis stems from the fact that when a real-world event occurs it usually has peak times on the Web: a higher volume of tweets, new visits and edits to related Wikipedia articles, and news published about the event. We propose a novel time-aware ranking model that leverages on multiple sources of crowd signals and integrates temporal signals in a learning to rank framework to rank results according to the predicted temporal relevance.
Our approach builds on two major novelties. First, a unifying approach that given a query q represents its temporal evidence mined from multiple sources of crowd signals. This allows us to predict the temporal relevance of documents for query q. Second, a principled retrieval model that integrates temporal signals in a learning to rank framework, to rank results according to the predicted temporal relevance. Evaluation on the TREC 2013 and 2014 Microblog track datasets demonstrates that the proposed model achieves a relative improvement of 13.2% over lexical retrieval models and 6.2% over a learning to rank baseline.
This presentation is based on a paper accepted to WSDM 2016.
Bio:
Flávio Martins is a Ph.D. candidate working with the group of Web and Media Search (NOVA Search). He is co-advised by João Magalhães in the NOVA Laboratory for Computer Science and Informatics (NOVA LINCS) at Universidade NOVA de Lisboa (UNL), and Jamie Callan in the Language Technologies Institude (LTI) at Carnegie Mellon University (CMU). He holds a M.Sc. in Computer Science from Universidade NOVA de Lisboa (UNL). His research interests include information retrieval and machine learning with a focus on real-time social-media search and time-aware retrieval. As part of his dissertation work he studies and builds techniques for learning-to-rank using temporal signals from multiple sources on the Web to improve search. Previously, he has worked in information retrieval techniques and systems in the medical domain as well as for federated search on the Web. He has already served in the Program Committees of the prestigious conferences ECIR and ACM CIKM and was a local organizer for the 10th OAIR conference.