What is it about?
This research introduces a new way for computers to recommend optimal teams of experts for complex projects. Instead of treating team formation as a simple checklist where each expert is selected independently, we reframe it as a translation problem - converting a list of required skills into a sequence of recommended experts. We tested this approach using sequence-to-sequence neural networks (similar to those used in language translation) on four large datasets from academic publications, patents, movies, and software development. The results show that our translation-based approach consistently outperforms traditional methods, achieving up to 82 times better performance in some cases. This breakthrough enables more effective team recommendations across various domains, from research collaborations to software development teams.
Featured Image
Photo by Pawel Czerwinski on Unsplash
Why is it important?
This work represents a fundamental shift in how we approach automated team recommendation. Traditional neural methods suffer from the "curse of sparsity" - with thousands of potential experts, the mathematical models struggle to learn meaningful patterns when most experts aren't selected for any given team. Our sequence-based approach solves this by treating team formation as a sequential decision process, where each expert selection influences the next. This mirrors how teams are actually formed in practice - you don't select members independently, but consider how they'll work together. The approach's success across diverse domains (academia, patents, entertainment, and software) demonstrates its broad applicability. As collaborative projects become increasingly complex and global, having effective automated team recommendation systems becomes crucial for organizational success and innovation.
Perspectives
Working on this paper was particularly exciting because it challenged a fundamental assumption in the field - that team recommendation should be treated as a classification problem. The inspiration came from observing how machine translation had evolved from word-by-word translation to sophisticated sequence models that understand context. We realized the same principle could apply to team formation. The most surprising finding was the magnitude of improvement - in some cases, our approach performed 82 times better than existing methods. This wasn't just an incremental improvement but a paradigm shift. We're especially proud that this work opens up new research directions, as the team recommendation community can now leverage decades of advances in sequence modelling. Looking forward, we hope this inspires researchers to reconsider other recommendation problems that might benefit from sequence-based approaches rather than traditional classification methods.
Kap Thang
University of Windsor
Read the Original
This page is a summary of: Translative Neural Team Recommendation: From Multilabel Classification to Sequence Prediction, July 2025, ACM (Association for Computing Machinery),
DOI: 10.1145/3726302.3730259.
You can read the full text:
Contributors
The following have contributed to this page
