My main research interests are in Knowledge Engineering, and more particularly in conceptual representation and inference processes applied to problem-solving.
I am currently exploring the synergies existing between Semantic Technologies and Data Mining, interesting since the need for « explainability » for learning algorithms has increased. Indeed, we are overwhelmed by huge amounts of information coming from different sources of great diversity, often associated with the semantic heterogeneity between the underlying conceptual models. Current approaches to Big Data processing are fundamentally concerned with the use of statistical techniques for understanding these data that are generated at high spatial, temporal or thematic resolutions. These approaches assume that the data are available. However, they do not answer an important set of questions concerning the retrieval of the data of interest, or the effective publication of these data, or the exploration of unknown datasets in different domains, or the correct understanding and interpretation of the data, among others. Semantic technologies, based entirely on open and well-established standards, can help in the integration of these heterogeneous data because they are capable of answering these questions.
Keywords: knowledge engineering, conceptualisation, ontologies and formal models, rule-based (crisp, fuzzy, probabilistic, spatio-temporal) reasoning, case‐based reasoning, knowledge and experience capitalisation.
The projects I am currently working on are:
- RESPONdING – REcherche Sémantique sur un corPus dOcumentaire pour la coNception en INGénierie
The main objective of these works is to qualitatively improve the search engine of our industrial partner, so that users can find, with the least amount of interaction possible, the documents relevant to their query and thus, the documents they will download. This will be done by improving and adding functionalities concerning the assistance to queryin, the relevance of the selected documents, their classification and the recommendations that the system can make. We are moving from a parading « querying over words » towards a new paradigm « querying over concepts ». - SA4XAI- Symbolic Approaches for Explainable Artificial Intelligence
For artificial intelligence (AI) systems to be understandable, symbolic systems and reasoning must be integrated to provide operationally effective communications of the internal state and the operation of the system.
Meaningful ontologies, knowledge bases and symbolic AI methods are fundamental (but currently, mostly ignored) to the most important explainable AI use-case: when in practice, within some context, a final user must understand, trust, and be responsible for the conclusions an AI system draws.
Thus, the main goal of this work is to explore the possible contributions of semantics and symbolic reasoning to take a step towards a more explainable artificial intelligence. - XAI for Manufacturing
This is a binational project (with the Department of Computer Science, ITU Norges Arktiske Universitet, in Norway), whose aim is to explore various aspects of the concept of explainability of artificial intelligence systems applied to production and Industry 4.0. These works will complete the results of the STEaMINg (Semantic Time Evolving Models for Industry 4.0) described with my previous projects and will be enriched with the results of the SA4XAI project.
My former research projects can be seen here.