Continuous spatial keyword search with query result diversifications

With the proliferation of location-based social media, it is of great importance to provide web users and mobile users with timely and high-quality information. In this light, we study the problem of continuous spatial keyword search over a stream of spatio-temporal messages by taking spatial relevance, textual relevance, and result diversification into consideration. We define a novel continuous query named Diversified Continuous Spatial Keyword (DCSK) query. A DCSK query consists of a spatial region, a set of query keywords, a similarity threshold 𝜃 , and the number of query results k . Given a stream of spatio-temporal messages, the DCSK query continuously receive spatio-temporal messages such that: (1) they are located inside the query region and contain at least one query keyword, and (2) the similarities between each message and its previous k messages are all mess than 𝜃 . Compared to traditional continuous spatial keyword query, the DCSK query can provide subscribers with spatio-temporal messages of higher quality because that the DCSK query takes both spatio-temporal relevance and query result diversification into consideration. We develop a Spatio-temporal Diversified Publish/Subscribe (STD-PS) framework to process a large number of DCSK queries efficiently. We conduct extensive experiments with real-world datasets. Our experimental results confirm the capability of our proposal in terms of result diversity, efficiency, and salability.