Finding the perfect match between a job proposal and a set of freelancers is not an easy task to perform at scale, especially in multiple languages. In this paper, we propose a novel neural retriever architecture that tackles this problem in a multilingual setting. Our method encodes project descriptions and freelancer profiles by leveraging pre-trained multilingual language models. The latter are used as backbone for a custom transformer architecture that aims to keep the structure of the profiles and project. This model is trained with a contrastive loss on historical data. Thanks to several experiments, we show that this approach effectively captures skill matching similarity and facilitates efficient matching, outperforming traditional methods.
@article{arxiv.2409.12097,
title = {Skill matching at scale: freelancer-project alignment for efficient multilingual candidate retrieval},
author = {Warren Jouanneau and Marc Palyart and Emma Jouffroy},
journal= {arXiv preprint arXiv:2409.12097},
year = {2024}
}