Improving Maps Auto-Complete Through Query Expansion (Demo Paper)

Abstract

Maps Auto-complete is an essential service complementing the functionality of map search engines. It allows users to formulate their queries faster and also provides better query formatting, which increases the chance of returning a relevant search result. Intuitively, the engagement with the service depends primarily on the quality of the suggestions it recommends. We notice, however, an interesting phenomenon that has not received much attention previously - often Auto-complete correctly identifies the most relevant suggestion, yet users do not click on it right away, if at all. Here we reason over the causes for the phenomenon, provide empirical evidence, and then propose a mitigation based on query expansion. Two models are proposed which generate word or phrase query expansions, allowing users to reach faster a 'mental pause' during which they are more likely to engage with the Auto-complete suggestions. Evaluation of the models is presented.