Enhancing Task Matching in Online Labor Markets Using Multi-field Features Interaction and Meta-learning

The online labor markets have facilitated the growth of customized services through digital platforms that connect employers with workers. However, the limited interactive information between workers and tasks has led to a “cold-start problem”, which limits the effectiveness of personalized task recommendation systems. To address this challenge, we purposed a personalized recommendation system for task recommendation, which combines the multi-field feature interaction and meta-learning. Our approach aims to lcapture concealed associations between multi-field features derived from both workers and tasks, thereby obtaining more meaningful worker preferences. Moreover, it captures workers personalized preferences with minimal interactions via meta learning, significantly enhancing cold-start recommendation performance. We evaluate the proposed model on a real-world dataset obtained from Freelancer.com, and the results demonstrate its superiority over three benchmark methods. By matching tasks with the most suitable workers, our system has the potential to reduce task completion times and enhance overall task quality.