# Learning Student Interest Trajectory for MOOCThread Recommendation

**Authors:** Shalini Pandey, Andrew Lan, George Karypis, Jaideep Srivastava  
**arXiv:** [2101.05625](https://arxiv.org/abs/2101.05625)  
**Date:** 2021-01-10  
**Summary page:** https://www.summarizepaper.com/en/paper/2101.05625v2/

## Key points

- Massive Open Online Courses (MOOCs) have become increasingly popular in recent years.
- The Covid-19 pandemic has further accelerated the growth of MOOCs and online education.
- Discussion forums are a primary means of interaction among learners and instructors in MOOCs.
- Students face challenges in finding useful and informative discussion forums as class sizes grow.
- Matching student interests with thread contents is crucial to address this problem.
- Authors propose a solution by predicting future interest trajectories of students.
- The model consists of two key operations: Update operation and Projection operation.
- Update operation captures how student interests are influenced by their interactions with threads.
- Projection operation estimates future embeddings of students and threads, adapting to changes in course topics and post structures.
- Extensive experimentation shows that the authors' model outperforms other baselines for thread recommendation significantly.
- Accurately predicting future interest trajectories enhances the learning experience in MOOCs.

## Summary

In recent years, Massive Open Online Courses (MOOCs) have experienced a surge in popularity. This growth has been further accelerated by the Covid-19 pandemic, which has necessitated the exploration of online education to its fullest potential. One of the primary means of interaction among learners and instructors in MOOCs is through discussion forums. However, as class sizes continue to grow, students are faced with the challenge of finding useful and informative discussion forums. To address this problem, it is crucial to match the interests of students with the contents of threads. The main challenge lies in the fact that student interests tend to change as they progress through a course, while forum contents evolve as students or instructors update them. In their paper titled "Learning Student Interest Trajectory for MOOCThread Recommendation," authors Shalini Pandey, Andrew Lan, George Karypis, and Jaideep Srivastava propose a solution by predicting future interest trajectories of students. The authors' model consists of two key operations: the Update operation and the Projection operation. The Update operation models the inter-dependency between the evolution of students and threads using coupled Recurrent Neural Networks when a student posts on a thread. This operation captures how student interests are influenced by their interactions with specific threads. On the other hand, the Projection operation focuses on estimating future embeddings of both students and threads. For students, the Projection operation learns to identify changes in their interests caused by shifts in course topics they study. By understanding these shifts, recommendations can be tailored accordingly to ensure continued engagement and relevance for each student. For threads, this operation leverages how different posts induce varying interest levels in a student based on thread structures. Extensive experimentation conducted on three real-world MOOC datasets demonstrates that the authors' model outperforms other baselines for thread recommendation significantly. By accurately predicting future interest trajectories and effectively matching them with relevant thread contents, this model offers an enhanced learning experience for MOOC participants. In conclusion, this research contributes to pushing the boundaries of online education and improving the overall learning experience in MOOCs.