Fabrice Popineau

GLAM: a generic layered adaptation model for adaptive hypermedia systems

This paper introduces GLAM, a system based on situation calculus and meta-rules, which is able to provide adaptation by means of selection of actions. It is primarily designed to provide adaptive navigation. The different levels of conception, related to different aspects of the available metadata, are split in different layers in GLAM, in order to ease the conception of the adaptation system as well as to increase the potential use of complex adaptation mechanisms. GLAM uses meta-rules to handle these layers.

Situation calculus and personalized web systems

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

Unsupervised Approach for Identifying Users' Political Orientations

Opinions, in news media platforms, provide a world wide access to what people think about daily life topics. Thus, exploiting such a source of information to identify the trends can be very useful in many scenarios, such as political parties who are interested in monitoring their impact. In this paper, we present an unsupervised technique to classify users based on their political orientations. Our approach is based on two main concepts: 1 the selection of the aspects and the sentiments users have expressed in their opinions, and 2 the creation of knowledge base from Wikipedia to automatically classify users according to their political orientations. We have tested our approach on two datasets crawled from CNN and Aljazeera. The results show that our approach achieves high quality results.

Adaptive Testing Using a General Diagnostic Model

In online learning platforms such as MOOCs, computerized assessment needs to be optimized in order to prevent boredom and dropout of learners. Indeed, they should spend as little time as possible in tests and still receive valuable feedback. It is actually possible to reduce the number of questions for the same accuracy with computerized adaptive testing (CAT): asking the next question according to the past performance of the examinee. CAT algorithms are divided in two categories: summative CATs, that measure the level of examinees, and formative CATs, that provide feedback to the examinees at the end of the test by specifying which knowledge components need further work. In this paper, we formalize the problem of test-size reduction by predicting student performance, and propose a new hybrid CAT algorithm GenMA based on the general diagnostic model, that is both summative and formative. Using real datasets, we compare our model to popular CAT models and show that GenMA achieves better accuracy while using fewer questions than the existing models.

Parametric Reasoning Agents Extend the Control over Standard Behaviors

In this paper, we study how to extend the control over the behavioral space of logically specified agents so as they can take into account parameters issuing from constraints and/or preferences that are external to their core reasoning process. Our approach is supported by the proposition of a set of generic transformations to apply to the original agent program that is expressed in Indigo log, a variant of the Golog language. Several of these transformations exploit the non-determinism present in Indigo log programs. We also propose an automatic process to apply all those transformations on the basis of a set of parameters. Three case-studies show the significance of the approach in situations where agents are in interaction with human users.

A Formal Approach to Personalization

Personalized systems are a response to the increasing number of resources on the Internet. In order to facilitate their design and creation, we aim at formalizing them. In this paper, we consider the relationship between a personalized application and its non-personalized counterpart. We argue that a personalized application is a formal extension of a non-personalized one. We aim at characterizing the syntactic differences between the expression of the personalized and non-personalized versions of the application. Situation calculus is our framework to formalize applications. We introduce two scenarios of non-personalized application that we personalize to illustrate our approach.

GEAHS: A generic educational adaptive hypermedia system based on situation calculus

GEAHS is a platform designed to ease the development of Adaptive Educational Hypermedia, using standard formalisms. In this document, we explain the underlying principles of this platform. Genericity is achieved thanks to an adaptation engine based on situation calculus and RDF. This paper describes the main aspects of our system, as well as the use we make of situation calculus to create a simpler, more reusable adaptive hypermedia system.

Automated Test Assembly for Handling Learner Cold-Start in Large-Scale Assessments

In large-scale assessments such as the ones encountered in MOOCs, a lot of usage data is available because of the number of learners involved. Newcomers, that just arrive on a MOOC, have various backgrounds in terms of knowledge, but the platform hardly knows anything about them. Therefore, it is crucial to elicit their knowledge fast, in order to personalize their learning experience. Such a problem has been called learner cold-start. We present in this article an algorithm for sampling a group of initial, diverse questions for a newcomer, based on a method recently used in machine learning: determinantal point processes. We show, using real data, that our method outperforms existing techniques such as uncertainty sampling, and can provide useful feedback to the learner over their strong and weak points.

Towards better news article recommendation

News media platforms publish articles about daily events letting their users comment on them, and forming interesting discussions in almost real-time. To keep users always active and interested, media platforms need an effective recommender system to bring up new articles that match user interests. In this article, we show that we can improve the quality of recommendation by exploiting valuable information provided by user comments. This information reveals aspects not directly tackled by the news article on which they have been posted. We call such aspects latent aspects. We demonstrate how these latent aspects can make a crucial difference in the accuracy of future recommendation. The challenge in detecting them is due to the noisy nature of user comments. To support our claim, we propose a novel news recommendation system that (1) enriches the description of news articles by latent aspects extracted from user comments, (2) deals with noisy comments by proposing a model for user comments ranking, and (3) proposes a diversification model to remove redundancies and provide a wide coverage of aspects. We have tested our approach using large collections of real user activities in four news Web sites, namely The INDEPENDENT, The Telegraph, CNN and Al-Jazeera. The results show that our approach outperforms baseline approaches achieving a significantly higher accuracy.

A Review of Recent Advances in Adaptive Assessment

Computerized assessments are an increasingly popular way to evaluate students. They need to be optimized so that students can receive an accurate evaluation in as little time as possible. Such optimization is possible through learning analytics and computerized adaptive tests (CATs): the next question is then chosen according to the previous responses of the student, thereby making assessment more efficient. Using the data collected from previous students in non-adaptive tests, it is thus possible to provide formative adaptive tests to new students by telling them what to do next. This chapter reviews several models of CATs found in various fields, together with their main characteristics. We then compare these models empirically on real data. We conclude with a discussion of future research directions for computerized assessments.