ExpressIF® is a software suite that allows the production of explainable artificial intelligences based on reasoning and knowledge. It is the product of classical algorithms, the most recent state-of-the-art, and an efficient implementation.
Over the years, various contributors have brought their bricks to the building in order to bring you these advanced features.
Today, our research is mainly focused on new features to make it easier to extract knowledge from data, to improve the usability of ExpressIF® and the confidence you can have in its decisions.
Below, we cite a few lines of research, which are at relatively low TRLs but which you can access in collaborative projects.
We always model more operators so that knowledge is expressed in a form as close as possible to natural language. Whether in the temporal, spatial, space-time domain, or many others, we make sure that all your knowledge is expressible in ExpressIF®.
Beyond the state-of-the-art, we are developing innovative algorithms to extract knowledge from data of various types : numerical, images, spectra, signals. Exploiting the richness of the vocabulary that can be used by ExpressIF®, the models obtained are more easily interpreted.
We also try to address two classic problems: reducing the efficiency / interpretability dilemma and learning in the presence of little or a lot of data.
We are interested in questions of interpretability and explainability. Depending on the domain, we set up tools to allow the understanding of the models and the formulation of an explanation of the decisions.
We are also working on the form that these clues can take, in particular the textual form in order to produce a justification in natural language of the decision.