Spaced repetition is among the most studied learning strategies in cognitive science literature. It consists temporally distributing exposure to an information so as improve long-term memorization. Providing students with adaptive and personalized distributed practice schedule would benefit more than just a generic scheduler. However, applicability of such schedulers seems be limited pure memorization, e.g. flashcards or foreign language learning. In this article, we first frame research...

In this paper, we introduce a new method of representation learning that aims to embed documents in stylometric space. Previous studies the field authorship analysis focused on feature engineering techniques order represent document styles and enhance model performance specific tasks. Instead, directly space by relying reference set authors intra-author consistency property which is one two components our definition writing style. The main intuition paper can define general from such that,...

Personalization in the field of Technology Enhanced Learning (TEL) is a topic that received lot concern by researchers. At same time, there growing amount Open Educational Resources (OER) indexed according to W3C standards. Relevant OERs can usefully complement contents delivered learner during an online course. Computing best offer at each point his course aspect personalization we address this paper. We designed our MORS system solve problem context Massive Online Courses (MOOC). Our...

In large-scale assessments such as the ones encountered in MOOCs, a lot of usage data is available because number learners involved. Newcomers, that just arrive on MOOC, have various backgrounds terms knowledge, but platform hardly knows anything about them. Therefore, it crucial to elicit their knowledge fast, order personalize learning experience. Such problem has been called learner cold-start. We present this article an algorithm for sampling group initial, diverse questions newcomer,...

In formative assessments, one wants to provide a useful feedback the examinee at end of test. order reduce number questions asked in an assessment, adaptive testing models have been developed for cognitive diagnosis, such as ones encountered knowledge space theory. However, when skills assessed is very huge, methods cannot scale. this paper, we present new method tests and examinee, even with large databases skills. It will be used Pix, platform certification digital competencies every...

As with many other tasks, neural networks prove very effective for anomaly detection purposes. However, few deep-learning models are suited detecting anomalies on tabular datasets. This paper proposes a novel methodology to flag based TracIn, an influence measure initially introduced explicability The proposed methods can serve augment any unsupervised deep method. We test our approach using Variational Autoencoders and show that the average of subsample training points point as proxy...

Personalized systems are a response to the increasing number of resources on Internet, but can be difficult create. In order facilitate design and creation such personalized systems, we aim at formalizing them. The situation calculus is logical framework that has often been proposed model web applications even ones. However, details its use much more rarely explained. this paper will show it needed carefully consider which variant choose. We precisely why want so-called guarded action...

Deep learning for tabular data has garnered increasing attention in recent years, yet employing deep models structured remains challenging. While these excel with unstructured data, their efficacy been limited. Recent research introduced retrieval-augmented to address this gap, demonstrating promising results supervised tasks such as classification and regression. In work, we investigate using anomaly detection on data. We propose a reconstruction-based approach which transformer model...

Self-supervision is often used for pre-training to foster performance on a downstream task by constructing meaningful representations of samples. Self-supervised learning (SSL) generally involves generating different views the same sample and thus requires data augmentations that are challenging construct tabular data. This constitutes one main challenges self-supervision structured In present work, we propose novel augmentation-free SSL method Our approach, T-JEPA, relies Joint Embedding...

With the tremendous growth of published news articles, a key issue is how to help users find diverse and interesting stories. To this end, it crucial understand build accurate profiles for both and news articles. In paper, we define user profile based on (1) set entities she/he talked about in her/his comments (2) key-concepts related those which has expressed a viewpoint. The same information extracted from content each article create its profile. These are then matched purpose...