Authors:
Mouncef Naji
1
;
Maroua Masmoudi
2
;
Hajer Baazaoui Zghal
1
;
Chirine Ghedira Guegan
3
;
Vlado Stankovski
4
and
Dan Vodislav
1
Affiliations:
1
ETIS Labs, CY Cergy Paris University, ENSEA / CNRS, France
;
2
CY Tech, Pau, CY Cergy Paris University, France
;
3
Univ. Lyon, Université Jean-Moulin Lyon 3, LIRIS UMR5205, iaelyon School of Management, France
;
4
Faculty of Civil and Geodetic Engineering, University of Ljubljana, Ljubljana, Slovenia
Keyword(s):
Big Data, Mapping Maintenance, Data Integration, Ontology, Deep Learning, Classification.
Abstract:
In recent years, the number of data sources and the amount of generated data are increasing continuously. This voluminous data leads to several issues of storage capacities, data inconsistency, and difficulty of analysis. In the midst of all these difficulties, data integration techniques try to offer solutions to optimally face these problems. In addition, adding semantics to data integration solutions has proven its utility for tackling these difficulties, since it ensures semantic interoperability. In our work, which is placed in this context, we propose a semantic-based data integration and mapping maintenance approach with application to drugs domain. The contributions of our proposal deal with 1) a virtual semantic data integration and 2) an automated mapping maintenance based on deep learning techniques. The goal is to support the continuous and occasional data sources changes, which would highly affect the data integration. To this end, we focused mainly on managing metadata
change within an integrated structure, refereed to as mapping maintenance. Our deep learning models encapsulate both convolutional, and Long short-term memory networks. A prototype has been developed and performed on two use cases. The process is fully automated and the experiments show significant results compared to the state of the art.
(More)