Fuzzy semantic approach for data integration applied to risk in food: an example about the cold chain
Résumé
A preliminary step to risk in food assessment is to gather experimental data. During the Sym’Previus project, we have designed a system for the integration of experimental data on food microbiology. Data provided by industrial partners and data extracted from experimental research results published in the main scientific journals of the domain are stored into a database. For that, data are indexed by means of a predefined vocabulary, called ontology. An important characteristic of the data in the database is their incompleteness. To enlarge the number of answers given to a user query, we have built an extended querying system called MIEL, which allows the user to retrieve the nearest data to his selection criteria. In this paper we detail another solution to deal with data scarcity, which consists in searching data on the Web to complement the database. Data tables that contain, in general, a synthesis of experimental data published in the documents, are automatically extracted from documents found on the web. The data tables are then automatically annotated with terms of the ontology and stored into the database. Each term belonging to a given data table is associated with a fuzzy set defined by the set of nearest terms belonging to the ontology weighted by a similarity score. Two complementary scores, which take into account “semantic” information provided by the ontology, are presented in this paper. This process, called fuzzy semantic annotation, permits the MIEL user to query the database, including data tables retrieved from the Web, using the same vocabulary (the ontology). We present an example of a query which can be useful in a cold chain study concerning E. coli O157 on ground beef.