Managing Web and social data

Managing Geo-RDF data

Description

Despite the large volume of work on querying large RDF knowledge bases, only a few studies efficiently address spatial semantics in RDF data; existing initiatives mainly focus on supporting GeoSPARQL features with external spatial indices like the R-tree, and less on performance optimization. Our work proposed a number of techniques that can be applied to RDF engines in order to efficiently support spatial queries such as range selections and spatial joins. As a proof of concept, the proposed techniques were implemented inside the RDF-3X open-source store. Our key contribution of is an effective encoding scheme that encodes spatial approximations inside the identifiers of RDF entities with spatial locations and geometries. This scheme allows us to apply online cheap filters and spatial join algorithms without accessing the exact geometries of the involved entities.

 

Publications

 

Analyzing Geo-social data

Description

The growth of Location-aware Social Networks (LASN) is unprecedented; given that users in LASNs such as Foursquare can rate venues, make recommendations and advertise interesting events, the economic importance and the potential impact of LASNs for businesses become apparent. Motivated by word-of-mouth and viral marketing campaigns, our work introduced the task of identifying the top-k Regionally Influential LASN Users (k-RIL). For example, an organizer of a city-wide festival is recruiting the most influential users within each district to locally advertise the festival. To solve k-RIL, we first proposed a novel propagation model and designed a methodology for computing the influence power of a user based on closeness centrality at the social graph. Then, we devised evaluation algorithms that examine candidate users in the order of their expected influence.

 

Publications

 

Managing Web directories

Description

Hierarchical structures such as topic directories are a popular way to organize information on the Web. In this context, we investigated the role of paths as knowledge artifacts; in particular, we proposed models to represent topic directories and a query language to manipulate path-like patterns together with raw data. Further, we studied the personlization of Web directories by mining user navigation patterns. For this purpose, we presented techniques for discovering group of users who exhibit similar navigation behaviour, termed interest groups. Essentially, the proposed personalization tasks aim at creating links termed shortcuts, between categories in the directory.

 

Publications