Applying MapReduce principle to high level information fusion

The InSyTo Synthesis framework is based on graph structures, graph algorithms and similarity measures for soft data fusion managing inconsistencies. The framework can be used to enable non-redundant additions to an information network, as well as graph based information query on several applications...

Full description

Saved in:
Bibliographic Details
Main Authors: Laudy, Claire, Dreo, Johann, Gouguenheim, Christophe
Format: Conference Proceeding
Language:English
Subjects:
Big data
Context
Data integration
Data models
Semantics
Terrorism
Vocabulary
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The InSyTo Synthesis framework is based on graph structures, graph algorithms and similarity measures for soft data fusion managing inconsistencies. The framework can be used to enable non-redundant additions to an information network, as well as graph based information query on several applications. The graph fusion algorithm relies on the search of a maximum common subgraph isomorphism, which makes it a difficult problem, especially on large graphs. In this work, the subgraph matching algorithm is partially parallelized, based on the MapReduce approach and on the Hadoop framework. Using Hadoop enables the management of big graphs, first by avoiding the load of the graphs in memory and secondly by distributing the computations over several processing nodes. Our experiments on the Global Terrorism Database (which contains the descriptions of more than 113,000 terrorist attacks in a graph of more than 20,000,000 nodes) shows that InSyTo Synthesis now scales to so-called "big data" applications.