The paper "Learning Dynamic Collective Entity Representations for Entity Ranking" is a cooperation between authors from the University of Amsterdam (David Graus, Maarten de Rijke), Yahoo! Research (Edgar Meij), and 904Labs.
In this paper, we look at entity ranking, i.e., successfully positioning a relevant entity at the top of the ranking for a given query. This task is inherently difficult due to the potential mismatch between the entity's description in a knowledge base (e.g., Wikipedia), and the way people refer to the entity when searching for it. To counter this issue we propose a method for constructing dynamic collective entity representations. The method collects entity descriptions from a variety of sources and combines them into a single entity representation. To this end, we learn to weigh the content from different sources that are associated with an entity for optimal retrieval effectiveness.
Our method is able to add new descriptions in real time, and learn the best representation at set time intervals. This allows us to to capture the dynamics in how people search entities as time progresses. The experimental results show that incorporating dynamic description sources into dynamic collective entity representations improves retrieval effectiveness by 7% over a state-of-the-art learning to rank baseline.
The paper was presented at the Web Search and Data Mining (WSDM 2016) conference in San Fransisco, Feb 22-25, 2016. For more information, check David Graus' website.