Sequential Learner Modeling Using Multi-Relational...

Sequential Learner Modeling Using Multi-Relational... | AI Research

Key Takeaways

Sequential Learner Modeling Using Multi-Relational Graph Convolutional Networks This paper introduces MR-ConceptGCN, a new approach designed to improve how e...
User modeling is a critical task in a variety of personalized systems.
Recognizing their effectiveness in learning from graph-structured data, Graph Neural Networks (GNNs), particularly Graph Convolutional Networks (GCNs), are increasingly employed for user modeling.
However, existing approaches typically treat different relation types in a graph as homogeneous, limiting their ability to capture richer semantics and construct more informative user models.
While multi-relational GNNs (MR-GNNs) have been adopted for representation learning and recommendation, their application for user modeling remains unexplored.

Paper AbstractExpand

User modeling is a critical task in a variety of personalized systems. Recognizing their effectiveness in learning from graph-structured data, Graph Neural Networks (GNNs), particularly Graph Convolutional Networks (GCNs), are increasingly employed for user modeling. However, existing approaches typically treat different relation types in a graph as homogeneous, limiting their ability to capture richer semantics and construct more informative user models. While multi-relational GNNs (MR-GNNs) have been adopted for representation learning and recommendation, their application for user modeling remains unexplored. Moreover, existing GNN-based user modeling approaches ignore the user interaction sequence. To address these research gaps, in this work we propose MR-ConceptGCN, a novel fully unsupervised approach focused on concept-based sequential learner modeling using multi-relational GCNs (MR-GCNs). MR-ConceptGCN effecively combines Personal Knowledge Graphs (PKGs), MR-GCNs, and the pre-trained language model SBERT to obtain enhanced relation- and semantic-aware representations of the PKG items. The enriched embeddings of the knowledge concepts that a learner did not understand when interacting with learning materials in CourseMapper are then used to construct a sequential learner model that combines long-term and short-term learner interactions. We report the results of an online user study (n = 31), demonstrating the benefits of MR-ConceptGCN in terms of several important user-centric aspects including accuracy, usefulness, diversity, and satisfaction with an educational recommender system.

Sequential Learner Modeling Using Multi-Relational Graph Convolutional Networks
This paper introduces MR-ConceptGCN, a new approach designed to improve how educational systems understand a learner's knowledge state. While many existing systems treat all interactions as the same, this research recognizes that educational concepts are connected in complex ways—such as one concept being a prerequisite for another. By combining Personal Knowledge Graphs (PKGs) with Multi-Relational Graph Convolutional Networks (MR-GCNs), the authors create a model that tracks both long-term and short-term learning patterns to provide more accurate and personalized recommendations.

Capturing Complex Relationships

Most current learner modeling tools rely on a simplified view of how concepts relate to one another. The authors argue that this misses critical information. Their approach, MR-ConceptGCN, explicitly incorporates multiple types of relationships, specifically "RELATED_TO" and "PREREQUISITE_TO." By using MR-GCNs, the system can distinguish between these different types of connections, allowing it to build a more nuanced map of what a student knows and what they still need to learn.

Combining Structure and Meaning

To make the model as effective as possible, the researchers combine structural graph data with semantic information. They use a pre-trained language model called SBERT to measure the semantic similarity between concepts. During the learning process, the system scales the information passed between nodes in the graph based on this similarity. This ensures that the resulting "knowledge state" of the learner is informed by both the logical structure of the curriculum and the actual meaning of the educational materials.

Modeling Learning Over Time

A major limitation of previous models is that they often aggregate all student data into a single, static snapshot, ignoring the order in which learning happens. MR-ConceptGCN addresses this by focusing on the sequence of interactions. By tracking the concepts a learner struggles with—those marked as "Did Not Understand"—the model builds a sequential profile that accounts for how a student’s needs evolve over time, effectively balancing their long-term goals with their immediate, short-term learning challenges.

Proven Benefits for Students

The researchers tested their approach through an online study with 31 participants using the CourseMapper platform. The results showed that MR-ConceptGCN successfully improved the educational recommender system across several key areas. Students reported higher satisfaction and found the system to be more accurate, useful, and diverse in its recommendations compared to traditional methods. This suggests that incorporating multi-relational, sequence-aware modeling is a highly effective way to support personalized learning.