Top-k Temporal Keyword Query over Social Media Data

Fan Xia, Chengcheng Yu, Weining Qian, Aoying Zhou

Abstract: Analytic jobs over social media data typically need to explore data of different periods. However, most existing keyword search work merely use creation time of items as the measurement of their recency. In this paper we propose top-k temporal keyword query that ranks data by their aggregate sum of shared times during the given time window. A query algorithm that can be executed over a general temporal inverted index is provided. The complexity analysis based on the power law distribution reveals the upper bound of accessed items. Furthermore, two-tiers structure and piecewise maximum approximation sketch are proposed as refinements. Extensive empirical studies on a reallife dataset show the combination of two refinements achieves remarkable performance improvement under different query settings.