Machine learning has proven increasingly essential in many fields. Yet, a lot obstacles still hinder its use by non-experts. The lack of trust in the results obtained is foremost among them, and has inspired several explanatory approaches in the literature. In this paper, we are investigating the domain of single prediction explanation. This is performed by providing the user a detailed explanation of the attribute's influence on each single predicted instance, related to a particular machine learning model. A lot of possible explanation methods have been developed recently. Although, these approaches often require an important computation time in order to be efficient. That is why we are investigating about new proposals of explanation methods, aiming to increase time performances, for a small loss in accuracy.

No abstract

No abstract

Searching information can be realized thanks to specific tools called Information Retrieval Systems IRS (also called “search engines”). To provide more accurate results to users, most of such systems offer personalization features. To do this, each system models a user in order to adapt search results that will be displayed. In a multi-application context (e.g., when using several search engines for a unique query), personalization techniques can be considered as limited because the user model (also called profile) is incomplete since it does not exploit actions/queries coming from other search engines. So, sharing user models between several search engines is a challenge in order to provide more efficient personalization techniques. A semantic architecture for user profile interoperability is proposed to reach this goal. This architecture is also important because it can be used in many other contexts to share various resources models, for instance a document model, between applications. It is also ensuring the possibility for every system to keep its own representation of each resource while providing a solution to easily share it.

No abstract

An enterprise memory must be able to be used as a basis for the processes of scientific or technical developments. Indeed, it was proven that information useful to these processes is not found solely in the operational bases of companies; it is also found in textual information and exchanged documents. For that reason, we propose the design and implementation of a documentary memory for business document warehouses. Its main characteristic is to allow the storage, retrieval, interrogation and analysis of information extracted from disseminated sources and, in particular, from the Web.

No abstract

As Machine Learning (ML) is now widely applied in many domains, in both research and industry, an understanding of what is happening inside the black box is becoming a growing demand, especially by non-experts of these models. Several approaches had thus been developed to provide clear insights of a model prediction for a particular observation but at the cost of long computation time or restrictive hypothesis that does not fully take into account interaction between attributes. This paper provides methods based on the detection of relevant groups of attributes -named coalitions - influencing a prediction and compares them with the literature. Our results show that these coalitional methods are more efficient than existing ones such as SHapley Additive exPlanation ( SHAP ). Computation time is shortened while preserving an acceptable accuracy of individual prediction explanations. Therefore, this enables wider practical use of explanation methods to increase trust between developed ML models, end-users, and whoever impacted by any decision where these models played a role.