Juliette Dibie-Barthélemy

Fuzzy semantic tagging and flexible querying of XML documents extracted from the Web

The relational database model is widely used in real applications. We propose a way of complementing such a database with an XML data warehouse. The approach we propose is generic, and driven by a domain ontology. The XML data warehouse is built from data extracted from the Web, which are semantically tagged using terms belonging to the domain ontology. The semantic tagging is fuzzy, since, instead of tagging the values of the Web document with one value of the domain ontology, we propose to use tags expressed in terms of a possibility distribution representing a set of possible terms, each term being weighted by a possibility degree. The querying of the XML data warehouse is also fuzzy: the end-users can express their preferences by means of fuzzy selection criteria. We present our approach on a first application domain: predictive microbiology.

Fuzzy Annotation of Web Data Tables Driven by a Domain Ontology

We propose an automatic system for annotating accurately data tables extracted from the web. This system is designed to provide additional data to an existing querying system called MIEL, which relies on a common vocabulary used to query local relational databases. We will use the same vocabulary, translated into an OWL ontology, to annotate the tables. Our annotation system is unsupervised. It uses only the knowledge defined in the ontology to automatically annotate the entire content of tables, using an aggregation approach: first annotate cells, then columns, then relations between those columns. The annotations are fuzzy: instead of linking an element of the table with a precise concept of the ontology, the elements of the table are annotated with several concepts, associated with their relevance degree. Our annotation process has been validated experimentally on scientific domains (microbial risk in food, chemical risk in food) and a technical domain (aeronautics).

A semantic validation of conceptual graphs

The research works on knowledge validation aim at enhancing the quality of knowledge bases. The conceptual graph model is a knowledge representation model which belongs to the family of the semantic networks. We give a solution to validate semantically a knowledge base expressed in terms of conceptual graphs. The semantic validation of a knowledge base consists in checking that the knowledge base respects a set of constraints given by an expert. We propose to express these constraints in terms of conceptual graphs. Two categories of constraints are introduced: the existential constraints which enable one to represent pieces of knowledge that must or must not exist in the knowledge base and the descriptive constraints which enable one to describe how some pieces of knowledge must be represented in the knowledge base. The checking of these constraints by a knowledge base is done by means of the projection operation which is the ground operation of the conceptual graph model.

Fuzzy Web Data Tables Integration Guided by an Ontological and Terminological Resource

In this paper, we present the design of ONDINE system which allows the loading and the querying of a data warehouse opened on the Web, guided by an Ontological and Terminological Resource (OTR). The data warehouse, composed of data tables extracted from Web documents, has been built to supplement existing local data sources. First, we present the main steps of our semiautomatic method to annotate data tables driven by an OTR. The output of this method is an XML/RDF data warehouse composed of XML documents representing data tables with their fuzzy RDF annotations. We then present our flexible querying system which allows the local data sources and the data warehouse to be simultaneously and uniformly queried, using the OTR. This system relies on SPARQL and allows approximate answers to be retrieved by comparing preferences expressed as fuzzy sets with fuzzy RDF annotations.

Flexible SPARQL Querying of Web Data Tables Driven by an Ontology

This paper concerns the design of a workflow which permits to feed and query a data warehouse opened on the Web, driven by a domain ontology. This data warehouse has been built to enrich local data sources and is composed of data tables extracted from Web documents. We recall the main steps of our semi-automatic method to annotate Web data tables driven by a domain ontology. The output of this method is an XML/RDF data warehouse composed of XML documents representing Web data tables with their fuzzy RDF annotations. We then present how to query simultaneously the local data sources and the XML/RDF data warehouse, using the domain ontology, through a flexible querying language. This language allows preferences to be expressed in selection criteria using fuzzy sets. We study more precisely how to retrieve approximate answers extracted from the Web data tables by comparing preferences expressed as fuzzy sets with fuzzy annotations using SPARQL.

Flexible Querying of Fuzzy RDF Annotations Using Fuzzy Conceptual Graphs

This paper presents a flexible querying system of fuzzy RDF annotations which consists in translating fuzzy RDF annotations into fuzzy conceptual graphs and using an approximate-projection operation in order to compare fuzzy query graphs with fuzzy annotation graphs. The fuzzy sets in the query graphs having a semantic of preferences are compared with the fuzzy sets in the annotation graphs having a semantic of similarity or imprecision. These comparisons deliver several scores which are used by our flexible querying system to sort the answers according to a total order even if these scores are not commensurable.

Semantic annotation of Web data applied to risk in food

A preliminary step to risk in food assessment is the gathering of experimental data. In the framework of the SymPrevius project (http://www.symprevius.org), a complete data integration system has been designed, grouping data provided by industrial partners and data extracted from papers published in the main scientific journals of the domain. Those data have been classified by means of a predefined vocabulary, called ontology. Our aim is to complement the database with data extracted from the Web. In the framework of the WebContent project (www.webcontent.fr), we have designed a semi-automatic acquisition tool, called @WEB, which retrieves scientific documents from the Web. During the @WEB process, data tables are extracted from the documents and then annotated with the ontology. We focus on the data tables as they contain, in general, a synthesis of data published in the documents. In this paper, we explain how the columns of the data tables are automatically annotated with data types of the ontology and how the relations represented by the table are recognised. We also give the results of our experimentation to assess the quality of such an annotation.

An ontology-driven annotation of data tables

This paper deals with the integration of data extracted from the web into an existing data warehouse indexed by a domain ontology. We are specially interested in data tables extracted from scientific publications found on the web. We propose a way to annotate data tables from the web according to a given domain ontology. In this paper we present the different steps of our annotation process. The columns of a web data table are first segregated according to whether they represent numeric or symbolic data. Then, we annotate the numeric (resp.symbolic) columns with their corresponding numeric (resp. symbolic) type found in the ontology. Our approach combines different evidences from the column contents and from the column title to find the best corresponding type in the ontology. The relations represented by the web data table are recognized using both the table title and the types of the columns that were previously annotated. We give experimental results of our annotation process, our application domain being food microbiology.

Flexible querying of Web data to simulate bacterial growth in food.

A preliminary step in microbial risk assessment in foods is the gathering of experimental data. In the framework of the SymPrevius project, we have designed a complete data integration system opened on the Web which allows a local database to be complemented by data extracted from the Web and annotated using a domain ontology. We focus on the Web data tables as they contain, in general, a synthesis of data published in the documents. We propose in this paper a flexible querying system using the domain ontology to scan simultaneously local and Web data, this in order to feed the predictive modeling tools available on the SymPrevius platform. Special attention is paid on the way fuzzy annotations associated with Web data are taken into account in the querying process, which is an important and original contribution of the proposed system.

Approximate querying of XML fuzzy data

The MIEL++ system integrates data expressed in two different formalisms: a relational database and an XML database. The XML database is filled with data semi-automatically retrieved from the Web, which have been semantically enriched according to the ontology used in the relational database. These data may be imprecise and represented as possibility distributions. The MIEL++ querying system scans the two databases simultaneously in a transparent way for the end-user. To scan the XML database, the MIEL query is translated into an XML tree query. In this paper, we propose to introduce flexibility into the query processing of the XML database, in order to take into account the imperfections due to the semantic enrichment of its data. This flexibility relies on fuzzy queries and query rewriting which consists in generating a set of approximate queries from an original query using three transformation techniques: deletion, renaming and insertion of query nodes.