Efficient Multi-View Maintenance in the Social Semantic Web


Published on

This poster describes an efficient approach to maintaining multiple views on large, evolving social networks.

The Social Semantic Web (SSW) refers to the mix of RDF data in web content, and social network data associated with those who posted that content. Applications to monitor the
SSW are becoming increasingly popular. For instance, marketers want to look for semantic patterns relating to the content of tweets and Facebook posts relating to their products. Such applications allow multiple users to specify patterns of interest, and monitor them in real-time as new data gets added to the web or to a social network. In this paper, we develop the concept of SSW view servers in which all of these types of applications can be simultaneously monitored from such servers. The patterns of interest are views. We show that a given set of views can be compiled in multiple possible ways to take advantage of common substructures, and de fine the concept of an optimal merge. We develop a very fast MultiView algorithm that scalably and efficiently maintains multiple subgraph views. We show that our algorithm is correct, study its complexity, and experimentally demonstrate that our algorithm can scalably handle updates
to hundreds of views on real-world SSW databases with up to 540M edges.

1 Like
  • Be the first to comment

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Efficient Multi-View Maintenance in the Social Semantic Web

  1. 1. Efficient Multi-View Maintenance in the Social Semantic Web 1 2 3 Views on Social Networks Multiple Views Merging Views  Query = Subgraph matching query •  On large social networks, multiple views are often IDEA:Very often, view queries have overlapping  View = Query for which the answer set is maintained concurrently. subgraph structures (bold arcs). If we can overlay themaintained as the social network database is •  Maintaining multiple views is very expensive, in different view queries such that these sharedupdated particular for rapidly changing databases. substructures can be matched jointly rather than publish •  e.g. Twitter has over 340 million tweets / day independently, we can save a lot of time. ?person ?doc Health topic Business   Social network updates are edge insertions (removals), as comments topic ?article1 topic Care Analytics hence graph merging has to center around the inserted so ci at publish topic references ?article2 Health ed ?author ?article Care expert publish ?msg2 references edge type. ?msg1 publish references tweet expert tweet tweet   Developed subgraph matching algorithm that can follows ?msg ?expert ?expert associated ?person ?other process merged view queries efficiently. 4 5 6 Example Merge Merge Optimality Optimal Merge publish topic publish expert ?v15 associated ?v13 ?v12   Many possible ways to merge query graphs. We ?v9 ?v8 topic comments follows associated topic ?v14 tweet references want high connected overlap which results in most topic Health topic Business Health topic ?v3 ?v16 savings at update time. Care ?v3 Analytics topic Care Business references ?v6 references Analytics publish   Define merged view score as the sum of edge expert topic ?a2 publish ?v5 references expert expert ?a2 tweet publish tweet publish ?v5 ?v6 overlaps. tweet expert references associated follows tweet ?a1 ?v4 ?v7 ?a1 associated ?v4 follows ?v7 publish   Finding the optimal view wrt the merged view follows references associated topic ?v8 Edges mapped by  1,  2 and  3 score is NP–hard. ?v11 Edges mapped by 1, 2 and 3 publish Edges mapped topic publish Edges mapped topic associated comments only by  1 Edges mapped   Our greedy view merging algorithm finds near ?v12 tweet only by 1 Edges mapped only by 2 publish only by  2 ?v9 ?v11 ?v10 Edges mapped only by  3 optimal views in practice. ?v13 references Edges mapped only by 3 LEGEND LEGEND 7 8 9 Experiments Mul2-View-Maintenance-Performance- Applications 900%% 100.0%% 800%% Outperforming- 95.0%% Maintaining multiple views jointly as a merged view Improvement-Compared our merged multi-view maintenance 700%% 600%% 90.0%%algorithm against standard independent view 500%% 85.0%% leads to significant improvements. 400%%maintenance. 300%% 80.0%%   477% faster than standard view maintenance 200%%  6 real world social network datasets with up to 540 100%% 75.0%%million edges. 0%% 70.0%% Applications include:   Monitoring social networks % % r% % % t%  Randomly generated 12,000 queries with varying ics n al be ck ku ro rn tu ys Fli Or En   Fraud, security applications, alerts ou Ph udegree of overlap and averaged results over 750 trials. Yo eJ Liv   Business Analytics  All algorithm implemented in Java on top of the COSI Performance improvement of the Multi-View   Knowledge Discoverygraph database middleware.   Caching frequently asked queries Maintenance algorithm on 6 different social networks Matthias Broecheler, Andrea Pugliese, and VS Subrahmanian