Paper Number
2458
Paper Type
Completed
Description
Person-job matching is a typical dynamic process with bilateral interactions between job seekers and jobs, along with sample imbalance issues. These characteristics pose significant challenges when designing an intelligent person-job match method. In this paper, we propose a novel process-oriented view of the person-job matching problem and formulate it as a multi-step multi-objective bilateral match learning problem. Our method combines profile features and historical sequential behaviors to learn the bilateral attributes and dynamic preferences, with multimodal data integrated through various attention mechanisms, such as the orthogonal multi-head and gated mechanisms. The method includes a sequence update module to learn the bilateral preferences and their updates sensitive to feedback. Furthermore, the multi-step constraint effectively solves the problem of imbalanced samples through partial relationships and information transmission between multi-objectives. Abundant experiments show that our method outperforms state-of-the-art methods in providing successful matches and improving recruitment efficiency.
Recommended Citation
Shi, Xiaowei; Wang, Cong; and Wei, Qiang, "Theory-driven Bilateral Dynamic Preference Learning for Person and Job Match: A Process-oriented Multi-step Multi-objective Method" (2023). ICIS 2023 Proceedings. 7.
https://aisel.aisnet.org/icis2023/dab_sc/dab_sc/7
Theory-driven Bilateral Dynamic Preference Learning for Person and Job Match: A Process-oriented Multi-step Multi-objective Method
Person-job matching is a typical dynamic process with bilateral interactions between job seekers and jobs, along with sample imbalance issues. These characteristics pose significant challenges when designing an intelligent person-job match method. In this paper, we propose a novel process-oriented view of the person-job matching problem and formulate it as a multi-step multi-objective bilateral match learning problem. Our method combines profile features and historical sequential behaviors to learn the bilateral attributes and dynamic preferences, with multimodal data integrated through various attention mechanisms, such as the orthogonal multi-head and gated mechanisms. The method includes a sequence update module to learn the bilateral preferences and their updates sensitive to feedback. Furthermore, the multi-step constraint effectively solves the problem of imbalanced samples through partial relationships and information transmission between multi-objectives. Abundant experiments show that our method outperforms state-of-the-art methods in providing successful matches and improving recruitment efficiency.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.
Comments
13-DataAnalytics