Guest Speakers/ Conférenciers invités :


·        For Conferences: 3 guest speakers

·        For Tutorials: 3 guest speakers


Talk Information: Conferences & Tutorials


Jean-Yves GRESSER, Vice Chairman, Black Forest Group Inc. New York, USA & Associé gérant GELM Entreprises, Paris, France.

Title:   The Key Role of Ontologies in Dealing with (Financial) Risks & Crises. (FR: Le rôle clé des ontologies dans le traitement des risques (financiers) & des crises).

Today’s World is currently facing complex and pervasive situations which may result in major crises or catastrophic events. This may happen in sectors as diverse as health, climate, environment, energy, transportation and finance not to say terrorism or wars. Stakeholders are many and speak diverse specialty or common languages.Before 2005 there was no ontology trying to cover the whole field of security. Available ontologies were dealing with specific areas. Since then there has been several converging attempts, including our own, to base the management of security on ontologies.

Still most current observation or analysis tools give us little grasp on actual situations. Their scope tends to be narrow or too IT oriented: they lack the ability to integrate new knowledge or know-how from different business horizons.

The first obstacle is the lack of ontologies in specific business areas. Take for example banking and insurance, where very few models exist. We were actually the first to coalesce formally knowledge from both sectors finance and security in a broad perspective. We did so in starting to develop our own top level ontologies in finance then merged them with top level models which could be derived from standard technical approaches of security. We then drilled down into the first models along several directions seemingly relevant to the current crises and problems.

Our work on crisis management indicates that while an ontology is primarily a model or an abstraction from which new concepts may emerge or knowledge be better represented, an ontology diagram can be used as a “roadmap” in the real world, thus providing instant benefits.


Mike THELWALL, Statistical Cybermetrics Research Group &
School of Technology, University of Wolverhampton, UK.
Title:   Data Mining Sentiment from the Social Web: Twitter vs. YouTube.

The rise of the social applications has filled the web with informal content written by citizens in blogs, twitter, social networks, discussion forums and elsewhere. This is exploited by market researchers to gain insights into product reception and advertising impact. For instance, many businesses deploy applications to search for customer reactions to their products and brands online every day. This provides a cheap and timely source of feedback, and, if combined with methods to automatically detect sentiment in text, can deliver powerful market intelligence, such rapid identification of any sudden increases in negative feelings towards a product. This talk will describe current methods for automatic sentiment detection in a challenging environment: informal text in social web sites. Such text typically ignores many of the rules of grammar and spelling, undermining the effectiveness of traditional sentiment analysis algorithms, but new algorithms are able to take advantage of deviations from language norms to detect sentiment in new ways. This talk will compare the prospects for sentiment detection in Twitter and YouTube using the results of some large-scale analyses.



Marie-Christine ROUSSET, Laboratoire d'Informatique de Grenoble (LIG) &  University of Grenoble (Joseph Fourrier), France.
Title:   Reasoning on Web Data Semantics.

Providing efficient and high-level services for integrating, querying and managing Web data raises many  difficult challenges, because data are becoming ubiquitous, multi-form, multi-source and musti-scale.
Data semantics is probably one of the keys for attacking those challenges in a principled way. A lot of effort has been done in the Semantic Web community for describing the semantics of information  through ontologies.
In this talk, we will show that description logics provide a good model for specifying ontologies over Web data (described in RDF), but that restrictions are necessary in order to obtain scalable algorithms for checking data consistency and answering conjonctive queries. We will show that the DL-Lite family has good properties for combining ontological reasoning and data management at large scale, and is then a good candidate for beeing a Semantic Web data model.



Mustapha NADI, Laboratoire d'Instrumentation Electronique de Nancy & Université Henri Poincaré-Nancy 1, France.
Title:   Interactions between disciplines in research and innovation.

The obvious need for interaction between disciplines:

In the light of climate conferences, like at Copenhagen in december 2009, the challenges to find solutions for green energy or climate protection is obviously impossible to solve by an isolated discipline. This evidence that it is a nessity for researchers to work beyond their field with other disciplines remains diffuse and poorly formalized concerning its implementation. The convergence and pooling of these differences may arise innovations, those that are real technological leaps, not gadgets for marketing. After the era of new technologies in the 90's, the Silicon Valley is now building the green economy. Other challenges for science and technology are working to respond to societal needs through the combination of different fields of knowledge. Biomedical research is another example that will be discussed in this conference throught its prospective projects.

The reality of collaborations between different research fields:

The reality of this union of scientific areas remains mostly wishful thinking, the reality of practicing interdisciplinarity being more difficult. The researcher is not explicitly formed or used to work in a real context of interaction with other disciplines. Empirical self-training is usually the only way to form themselves during a common project.

One can look at the essence of these interactions that several studies have already attempted to define. What interests us here is an approach of their impact on research and innovation. Are they inhibiting or promoting factors for innovation and research?



Stéphane CHAUDIRON, GERiiCO Laboratory & University of Lille 3 (France).
Title:   Studying Web Search Engines from User Perspectives.

The wider use of Web search engines brings us to reconsider the theoretical and methodological frameworks to grasp new information practices. The number and diversity of studies currently focusing on Web search engines give valuable insights regarding the importance of this activity, which has become a large socio-economic, political and cultural phenomenon. Even if, during the past years, academic studies essentially have focused on the professional uses of information access systems, they are now more concerned with the question of the everyday use of these systems, by a “wider public”, in an everyday life context. Coming from different disciplines such as information science, communication science, sociology, psychology or marketing, user-oriented studies are meaningful to understand the different contexts of use of electronic information access systems. With different goals and methodologies, they all have in common to capture the “user behavior” and the “information practices” at different levels, search tasks and strategies, personal information infrastructure, cognitive or social-organizational contexts. The conference will present an overview of some approaches which have played and still play for some of them a dominant role in the comprehension of information access process.


Bruno CREMILLEUX, GREYC laboratory & University of Caen, France.
Title: Discovering useful information in data mining

Extracting or mining knowledge from large amounts of data is the aim of the Knowledge Discovery from Databases (KDD), often also named "data mining". Pattern discovery is at the core of the extracted knowledge and the KDD processes. There are now many approaches to discover information such as the so-called local patterns, these approaches running from several kinds of data (e.g., 0/1 data, sequences, graphs). Nevertheless, the overwhelming number of produced patterns hinders their uses. Such massive output hampers the individual analysis performed by the end-users of data and a current challenge in KDD is to cope with the "pattern flooding which follows data flooding".

After a short introduction to data mining and the pattern discovery task, we will present the usual approaches to reduce the number of outputted local patterns: the interestingness measures, the constraint-based paradigm and the pattern condensed representations. In order to discover useful information, a further and key step is to take into account the interest of a pattern with respect to the other patterns. This new trend is a powerful way to produce patterns of a higher level and models from the data. We will develop the emergent domain of crossing data mining and Constraint Programming to propose generic approaches for modeling and mining n-ary patterns and global patterns. Integrating background knowledge is also a way to focus on useful patterns. We will present methods based on a cross-fertlization between data mining and natural language processing to discover few patterns leading to a actionable information for the linguists.