Adaptive Sequential Recommendation for Discussion Forums on MOOCs using Context Trees

Massive open online courses (MOOCs) have demonstrated growing popularity and rapid development in recent years. Discussion forums have become crucial components for students and instructors to widely exchange ideas and propagate knowledge. It is important to recommend helpful information from forums to students for the benefit of the learning process. However, students or instructors update discussion forums very often, and the student preferences over forum contents shift rapidly as a MOOC progresses. So, MOOC forum recommendations need to be adaptive to these evolving forum contents and drifting student interests. These frequent changes pose a challenge to most standard recommendation methods as they have difficulty adapting to new and drifting observations. We formalize the discussion forum recommendation problem as a sequence prediction problem. Then we compare different methods, including a new method called context tree (CT), which can be effectively applied to online sequential recommendation tasks. The results show that the CT recommender performs better than other methods for MOOCs forum recommendation task. We analyze the reasons for this and demonstrate that it is because of better adaptation to changes in the domain. This highlights the importance of considering the adaptation aspect when building recommender system with drifting preferences, as well as using machine learning in general.

From data to structures

Vasilis Kalofolias

Weighted undirected graphs are a simple, yet powerful way to encode structure in data. A first question we need to address regarding such graphs is how to use them effectively to enhance machine learning problems. A second but more important question is how to obtain such a graph directly from the data. These two questions lead the direction of this thesis. We first explore the use of graphs in matrix completion, with direct implications in recommendation systems. In such systems side information graphs arise naturally, by comparing user behavior and interactions, and by product profiling. Given incomplete object ratings by users, and known graphs for user and object similarities, we propose a new matrix completion model that exploits the graphs to achieve better recommendations. Our model is convex and is solved efficiently with an Alternating Direction Multipliers method (ADMM). It achieves better recommendations than standard convex matrix completion, especially when very few ratings are known a priori, while it is robust to the graph quality. Continuing with the second question, we are interested in learning a graph from data. Data is usually assumed to be smooth on a graph, an assumption associated with graph sparsity and connectivity. Using this assumption, we provide a general graph learning framework that is convex and easy to optimize. We classify graph learning according to the prior structure information in edge-local, node-local or global. Edge-local models are the cheapest computationally, learning each edge independently. Exponential decay graphs and

\epsilon

-neighborhood graphs, the most prevalent weighting schemes in the literature, are such models of our framework. We provide a new model that interpolates between these two, and achieves better performance in many settings. Node-local models take into account the neighborhood of each edge. While they require iterative algorithms, they are stronger models achieving better results. We provide a new node-local model for general graph learning under smoothness assumptions, that redefines the state of the art. We discuss the previous state of the art, two global models of our framework. We provide efficient primal-dual based algorithms for efficiently solving our model and the most prevalent of the previous ones. Our model performs best in many settings. Large scale applications require graphs that are easy to compute, with cost that scales gracefully with respect to the number of nodes, and do not need parameter tuning. We provide the first large scale graph learning solution, based on approximate nearest neighbors with cost of

\mathcal{O}(m\log(m))

for

m

nodes. Our solution automatically finds the needed parameters given the sparsity level needed. We finally explore learning graphs that change in time. In many applications graphs are not static, and data comes from different versions of them. In such cases there is no one graph that correctly summarizes connectivity, and we need to learn several graphs, one for each selected time window. Assuming that graphs change slowly, we provide two new models based on a known global and our own node-local one. Our convex time-varying-graph models can use any convex differentiable loss function for time-change penalization, have a cost linear to the number of time windows, and are easy to parallelize. The obtained graphs fit the data better and provide a good compromise between detail in time and computational cost.

EPFL2016

Semiautomatic Annotation of MOOC Forum Posts

Pierre Dillenbourg, Lukasz Kidzinski, Weizhe Liu

Massive online open courses' (MOOCs') students who use discussion forums have higher chances of finishing the course. However, little research has been conducted for understanding the underlying factors. One of the reasons which hinders the analysis is the amount of manual work required for annotating posts. In this paper, we use machine learning techniques to extrapolate small set of annotations to the whole forum. These annotations not only allow MOOC producers to summarize the state of the forum, but they also allow researchers to deeper understand the role of the forum in the learning process.

Springer-Verlag Singapore Pte Ltd2016

Variations in Pedagogical Design of Massive Open Online Courses (MOOCs) Across Disciplines

Stian Haklev

Given that few studies have formally examined pedagogical design considerations of Massive Online Open Courses (MOOCs), this study explored variations in the pedagogical design of six MOOCs offered at the University of Toronto, while considering disciplinary characteristics and expectations of each MOOC. Using a framework (Neumann et al., 2002) characterizing teaching and learning across categories of disciplines, three of the MOOCs represented social sciences and humanities, or “soft” MOOCs, while another three represented sciences, or “hard” MOOCS. We utilized a multicase study design for understanding differences and similarities across MOOCs regarding learning outcomes, assessment methods, interaction design, and curricular content. MOOC instructor interviews, MOOC curricular documents, and discussion forum data comprised the data set. Learning outcomes of the six MOOCs reflected broad cognitive competencies promoted in each MOOC, with the structure of curricular content following disciplinary expectations. The instructors of soft MOOCs adopted a spiral curriculum and created new content in response to learner contributions. Assessment methods in each MOOC aligned well with stated learning outcomes. In soft MOOCs, discussion and exposure to diverse perspectives were promoted while in hard MOOCs there was more emphasis on question and answer. This study shows disciplinary-informed variations in MOOC pedagogy, and highlights instructors’ strategies to foster disciplinary ways of knowing, skills, and practices within the parameters of a generic MOOC platform. Pedagogical approaches such as peer assessment bridged the disciplines. Suggestions for advancing research and practice related to MOOC pedagogy are also included.

Indiana Univ Press2017

From data to structures

Vasilis Kalofolias

\epsilon

\mathcal{O}(m\log(m))

for

m

EPFL2016