Integrating platforms through content-based graph representation learning

Abstract

The advancements in education technology have encouraged businesses to offer a wide range of educational resources aimed at teaching labor market-related skills. However, the ever-changing abundance of information in both course and job platforms makes it difficult for job seekers to identify the right learning resources for a particular job opportunity. As a result, it is important to integrate and analyze information from both platforms to recommend suitable courses and jobs to them. Meanwhile, integrating information from these distinct platforms for analysis purposes poses a challenge due to the presentation of data in free text. Previous studies have mainly focused on downstream tasks in one area and few have integrated both of them. Additionally, linking skills from these platforms through methods such as crowdsourcing or annotation mechanisms may not be reliable due to potential biases or inconsistencies. This study proposes a novel approach to integrate data from course and job advertising platforms using graph representation learning and embedding algorithms. The proposed approach predicts the relationship between the two platforms and has been tested in the form of link prediction tasks using several graph learning methods with RMSE scores below 0.1 and AUC scores between 70-75%. These results imply that our approach can be further refined and has the potential to be used in building recommendation systems that help address the problem of information overload among students and job seekers.

Keywords

• Courses
• Jobs
• Skills
• Graph neural networks
• Representation learning
• Embedding algorithm