Reproduce “Professional Network Matters: Connections Empower Person-Job Fit” on OpenResearch

About this paper

Hao Chen, Lun Du, Yuxuan Lu, Qiang Fu, Xu Chen, Shi Han, Yanbin Kang, Guangming Lu and 1 more

Researchers from Microsoft Research and LinkedIn Corporation, in collaboration with academic institutions, developed a two-stage graph neural network framework to integrate professional network information into Person-Job Fit models. This method, comprising WHIN pre-training and a Contextual Social Attention GNN, consistently surpassed baseline models across three real-world LinkedIn datasets, demonstrating a 2.37% AUC improvement over HGT on the Hybrid dataset.

Key insights

• Professional networks are a crucial, yet underutilized, source of information for enhancing Person-Job Fit, particularly beneficial for job seekers with incomplete profiles.
• Successfully leveraging professional network data requires intelligent mechanisms, such as job-specific attention and relevance-based message passing, to filter out 'social noise'.
• A two-stage approach, involving pre-training on a Workplace Heterogeneous Information Network (WHIN) and then applying a Contextual Social Attention GNN (CSAGNN), effectively captures and utilizes heterogeneous relational knowledge.

Summary by alphaXiv.