Nathalie Pernelle

ZooM: a nested Galois lattices-based system for conceptual clustering

This paper deals with the representation of multi-valued data by clustering them in a small number of classes organized in a hierarchy and described at an appropriate level of abstraction. The contribution of this paper is three fold. First, we investigate a partial order, namely nesting, relating Galois lattices. A nested Galois lattice is obtained by reducing (through projections) the original lattice. As a consequence it makes coarser the equivalence relations defined on extents and intents. Second we investigate the intensional and extensional aspects of the languages used in our system ZooM. In particular we discuss the notion of α-extension of terms of a class language £. We also present our most expressive language £3, close to a description logic, and which expresses optionality or/and multi-valuation of attributes. Finally, the nesting order between the Galois lattices corresponding to various languages and extensions is exploited in the interactive system ZooM. Typically a ZooM session starts from a propositional language £2 and a coarse view of the data (through α-extension). Then the user selects two ordered nodes in the lattice and ZooM constructs a fine-grained lattice between the antecedents of these nodes. So the general purpose of ZooM is to give a general view of concepts addressing a large data set, then focussing on part of this coarse taxonomy.

Incremental Ontology-Based Extraction and Alignment in Semi-structured Documents

SHIRI is an ontology-based system for integration of semi-structured documents related to a specific domain. The systems purpose is to allow users to access to relevant parts of documents as answers to their queries. SHIRI uses RDF/OWL for representation of resources and SPARQL for their querying. It relies on an automatic, unsupervised and ontology-driven approach for extraction, alignment and semantic annotation of tagged elements of documents. In this paper, we focus on the Extract-Align algorithm which exploits a set of named entity and term patterns to extract term candidates to be aligned with the ontology. It proceeds in an incremental manner in order to populate the ontology with terms describing instances of the domain and to reduce the access to extern resources such as Web. We experiment it on a HTML corpus related to call for papers in computer science and the results that we obtain are very promising. These results show how the incremental behaviour of Extract-Align algorithm enriches the ontology and the number of terms (or named entities) aligned directly with the ontology increases.

Classification rule learning for data linking

Many approaches have been defined to link data items automatically. Nevertheless, when data are numerous and when the schema is unknown, most of these approaches are too time-consuming. We propose an approach where classification rules are learnt thanks to a training set made of linked data. These classification rules can then be applied in order to classify data items and reduce the linking space i. e the space made of data item pairs that have to be compared. First experiments have been conducted on RDF data sets describing electronic products.

Supporting semantic search on heterogeneous semi-structured documents

This paper presents SHIRI-Querying, an approach for semantic search on semi-structured documents. We propose a solution to tackle incompleteness and imprecision of semantic annotations of semistructured documents at querying time. We particularly introduce three elementary reformulations that rely on the notion of aggregation and on the document structure. We present the Dynamic Reformulation and Execution of Queries algorithm (DREQ) which combines these elementary transformations to construct reformulated queries w.r.t. a defined order relation. Experiments on two real datasets show that these reformulations greatly increase the recall and that returned answers are effectively ranked according to their precision.

Automatic Construction and Refinement of a Class Hierarchy over Multi-valued Data

In many applications, it becomes crucial to help users to access to a huge amount of data by clustering them in a small number of classes described at an appropriate level of abstraction. In this paper, we present an approach based on the use of two languages of description of classes for the automatic clustering of multi-valued data. The first language of classes has a high power of abstraction and guides the construction of a lattice of classes covering the whole set of the data. The second language, more expressive and more precise, is the basis for the refinement of a part of the lattice that the user wants to focus on.

A semantic enrichment of data tables applied to food risk assessment

Our work deals with the automatic construction of domain specific data warehouses. Our application domain concerns microbiological risks in food products. The MIEL++ system [2], implemented during the SymPrevius project, is a tool based on a database containing experimental and industrial results about the behavior of pathogenic germs in food products. This database is incomplete by nature since the number of possible experiments is potentially infinite. Our work, developed within the e.dot project, presents a way of palliating that incompleteness by complementing the database with data automatically extracted from the Web. We propose to query these data through a mediated architecture based on a domain ontology. So, we need to make them compatible with the ontology. In the e.dot project [5], we exclusively focus on documents in Html or Pdf format which contain data tables. Data tables are very common presentation scheme to describe synthetic data in scientific articles. These tables are semantically enriched and we want this enrichment to be as automatic and flexible as possible. Thus, we have defined a Document Type Definition named SML (Semantic Markup Language) which can deal with additional or incomplete information in a semantic relation, ambiguities or possible interpretation errors. In this paper, we present this semantic enrichment step.

Controlled knowledge base enrichment from web documents

The Linked Open Data initiative brought more and more RDF data sources to be published on the Web. However, these data sources contain relatively little information compared to the documents available on the surface Web. Many annotation tools have been proposed in the last decade for the automatic construction and enrichment of knowledge bases. But, while noticeable advances are achieved for the extraction of concept instances, the extraction of semantic relations remains a challenging task when the structures and the vocabularies of the target documents are heterogeneous. In this paper, we propose a novel approach, called REISA, which allows to enrich RDF/OWL knowledge bases with semantic relations using semistructured documents annotated with concept instances. REISA produces weighted relation instances without exploiting lexico-syntactic or structure regularities in the documents. Neighbor domain entities in the annotated documents are used to generate the first sets of candidate relations according to the domain and range axioms defined in a domain ontology. The construction of these candidate sets relies on automated semantic controls performed with (i) the existing knowledge bases and (ii) the (inverse) functionality of the target relations. The weighting of the selected relation candidates is performed according to the neighborhood distance between the annotated domain entities in the document. Experiments on two real web datasets show that (i) REISA allows to extract semantic relationships with interesting precision values reaching 76,5% and that (ii) the weighting method is effective for ranking the relation candidates according to their precision.

On the Explanation of SameAs Statements using Argumentation

Due to the impressive growing of the LOD graph in the last years, assuring the quality of its content is becoming a very important issue. Thus, it is crucial to design techniques for supporting experts in validating facts and links in complex data sources. Here, we focus on identity links (sameAs) and apply argumentation semantics to (i) detect inconsistencies in sameAs statements and to (ii) explain them to the experts using dialogues. We formalize the framework, explaining its purposes. Finally we provide a promising preliminary evaluation and discuss on some interesting future directions we foresee.

Discovering keys in RDF/OWL dataset with KD2R

KD2R allows the automatic discovery of composite key constraints in RDF data sources that conform to a given ontology. We consider data sources for which the Unique Name Assumption is fulfilled. KD2R allows this discovery without having to scan all the data. Indeed, the proposed system looks for maximal non keys and derives minimal keys from this set of non keys. KD2R has been tested on several datasets available on the web of data and it has obtained promising results when the discovered keys are used to link data. In the demo, we will demonstrate the functionality of our tool and we will show on several datasets that the keys can be used in a datalinking tool.

Explaining Reference Reconciliation Decisions: A Coloured Petri Nets Based Approach

Data integration systems aims at facilitating the management of heterogeneous data sources. When huge amount of data have to be integrated, resorting to human validations is not possible. However, completely automatic integration methods may give rise to decision errors and to approximated results. Hence, such systems need explanation modules to enhance the user confidence in the integrated data. In this paper, we focus our study on reference reconciliation methods which compare data descriptions to decide whether they refer to the same real world entity. Numerical reference reconciliation methods that are global and ontology driven, exploit semantic knowledge to model the dependencies between similarities and to propagate them to other references. In order to explain the similarity scores and the reconciliation decisions obtained by such methods, we have developed an explanation model based on Coloured Petri Nets which provides graphical and comprehensive explanations to the user. This model allows to show the relevance of one decision, and to diagnose possible anomalies in the domain knowledge or in the similarity measures that are used.