This chapter presents an application in the field of food safety using an ontology-based data integration approach. An ontology is a vocabulary used to express the knowledge in a given domain of application. In this chapter, the ontology-based data integration approach permits to homogenize data sources which are heterogeneous in terms of structure and vocabulary. This approach is done in the framework of the Semantic Web, an international initiative which proposes annotating data sources using ontologies in order to manage them more efficiently. In this chapter, the authors explore three ways to integrate data according to a domain ontology: (1) a semantic annotation process to extend local data with Web data which have been semantically annotated according to a domain ontology, (2) a flexible querying system to query uniformly both local data and Web data and (3) an ontology alignment process to find correspondences between data from two sources indexed by distinct ontologies.
TopIntroduction
The aim of the data integration systems is to put together a large amount of data coming from multiple and independent sources. One of the main problems of the data integration is the data heterogeneity. It can come from the structure of the data, the vocabulary used to index the data and the format of the data. These characteristics are in general specific to each source of data. Their harmonization is necessary to integrate the data. Another problem of the data integration is the data rarity. Although this problem can seem paradoxical, it can be explained by the fact that the numerous available data are not necessarily pertinent for a given application domain (in food safety for instance). A third problem may also occur in data integration: the imprecision of data. This imprecision can be intrinsic to the data (for instance an interval of pH values) or can correspond to a pertinence degree associated with the data according to the application domain.
We have developed a system, called CARAT (Chronic & Acute Risk Assessment), to estimate the exposure of a given population of consumers to chemical contaminants which relies on two distinct data sources. Both sources contain information about food products. The first one, called CONTA source, contains measures of the level of chemical contamination for food products. The second one, called CONSO source, stores household purchases of food products. Both sources have been indexed using their own domain ontology, the CONTA ontology and the CONSO ontology, an ontology representing a vocabulary used to express the knowledge in a given application domain. The CARAT system is composed of two sub-systems (see Figure 1): a decision support system that uses statistical methods to compute the exposure of a given population of consumers to chemical contaminants (Buche, Soler & Tressou, 2006) and an ontology-based data integration system which feeds the decision support system with data about the chemical contamination and the consumption of food products. The data integration system is managed using a data warehouse approach: data sources provided by external partners are replicated locally and standardized using ETL (Extract, Transform, Load) technology.
In this chapter, we present the ontology-based data integration system which takes into account the three data integration problems presented above: data heterogeneity, data rarity and data imprecision. The ontology-based data integration system proposes three different ways to integrate data according to a domain ontology. The first one is a semantic annotation process which allows a local database (the CONTA local database), indexed by a domain ontology, to be extended with data that have been extracted from the Web and semantically annotated according to this domain ontology. The second one, which is an original contribution of this chapter, is a querying system which allows the semantically annotated Web data to be integrated with the local data through a uniform flexible querying system relying on a domain ontology (the CONTA ontology). The third one is an ontology alignment method relying on rules which allow correspondences to be found between objects of a source ontology (the CONSO ontology) and objects of a target ontology (the CONTA ontology) according to their characteristics and associated values. Those three ways to integrate data have been designed using the Semantic web approach, an international initiative, which proposes annotating data sources using ontologies in order to manage them more efficiently.
In this chapter, we first present the ontology-based data integration system. We then provide some background on the topic. Third, current projects and future trends are presented. We conclude this chapter in the last section.
TopThe Ontology-Based Data Integration System
This section describes the different construction steps of the ontology-based data integration system of the CARAT system. In the first section, we present the filling of its data sources. In the second section, we present its querying system. In the third section, we present the alignment between objects of its two data sources which are indexed by distinct ontologies.