This is a page that curates AI-related papers published worldwide. All content here is summarized using Google Gemini and operated on a non-profit basis. Copyright for each paper belongs to the authors and their institutions; please make sure to credit the source when sharing.
MHSNet:An MoE-based Hierarchical Semantic Representation Network for Accurate Duplicate Resume Detection with Large Language Model
Created by
Haebom
Author
Yu Li, Zulong Chen, Wenjian Xu, Hong Wen, Yipeng Yu, Man Lung Yiu, Yuyu Yin
Outline
This paper proposes MHSNet, a novel framework for detecting duplicates in resumes collected from third-party websites to maintain a company's talent pool. MHSNet fine-tunes BGE-M3 using contrastive learning and uses Mixture-of-Experts (MoE) to generate multi-layered (sparse and dense) representations of resumes to compute semantic similarity. A notable feature is its use of state-aware MoE to handle a variety of incomplete resumes. Experimental results demonstrate the effectiveness of MHSNet.
Takeaways, Limitations
•
Takeaways:
◦
It can contribute to improving the quality of third-party resumes and expanding the company's talent pool.
◦
We present an effective duplicate detection method for incomplete and heterogeneous resume data.
◦
We present a novel approach to generating multi-layered semantic representations by combining contrastive learning and MoE.
•
Limitations:
◦
The performance evaluation of the proposed MHSNet may be limited to a specific dataset. Additional experiments on diverse datasets are needed.
◦
Further research is needed on applicability and scalability in real-world business environments.
◦
Because of the high dependence on BGE-M3, analysis of performance changes when using other base models is necessary.
