[Paper Review] Enhancing Person-Job Fit for Talent Recruitment: An Ability-aware Neural Network Approach

Chuan Qin, Hengshu Zhu+5 more|arXiv (Cornell University)|Dec 21, 2018

Topic Modeling33 references60 citations

TL;DR

The paper introduces APJFNN, an end-to-end neural network that uses word-level representations and four ability-aware attention mechanisms to measure and interpret Person-Job Fit from job postings and resumes.

ABSTRACT

The wide spread use of online recruitment services has led to information explosion in the job market. As a result, the recruiters have to seek the intelligent ways for Person Job Fit, which is the bridge for adapting the right job seekers to the right positions. Existing studies on Person Job Fit have a focus on measuring the matching degree between the talent qualification and the job requirements mainly based on the manual inspection of human resource experts despite of the subjective, incomplete, and inefficient nature of the human judgement. To this end, in this paper, we propose a novel end to end Ability aware Person Job Fit Neural Network model, which has a goal of reducing the dependence on manual labour and can provide better interpretation about the fitting results. The key idea is to exploit the rich information available at abundant historical job application data. Specifically, we propose a word level semantic representation for both job requirements and job seekers' experiences based on Recurrent Neural Network. Along this line, four hierarchical ability aware attention strategies are designed to measure the different importance of job requirements for semantic representation, as well as measuring the different contribution of each job experience to a specific ability requirement. Finally, extensive experiments on a large scale real world data set clearly validate the effectiveness and interpretability of the APJFNN framework compared with several baselines.

Motivation & Objective

• Motivate the need for improved Person-Job Fit in online recruitment under information explosion.
• Propose an end-to-end neural framework that leverages historical recruitment data to reduce manual labeling.
• Learn word-level semantic representations and hierarchical ability-aware representations for job requirements and candidate experiences.
• Provide interpretable matching results via ability-aware attention mechanisms.
• Demonstrate superior performance over baselines on a large real-world recruitment dataset.

Proposed method

• Represent job requirements and candidate experiences with word-level BiLSTM encodings.
• Construct hierarchical ability-aware representations for both job postings and resumes.
• Apply four hierarchical attention mechanisms to highlight key words and abilities, and to relate experiences to requirements.
• Compute a joint Person-Job Fit representation using a fully-connected predictor and optimize with binary cross-entropy.
• Train using an Adam optimizer with specific initialization and dropout to prevent overfitting.
• Compare APJFNN against classical ML baselines and a simpler BPJFNN baseline on a real-world dataset.

Experimental results

Research questions

• RQ1 How can word-level and hierarchical ability-aware representations improve Person-Job Fit accuracy and interpretability?
• RQ2 Can an end-to-end neural approach outperform traditional feature-based models for matching jobs to candidates?
• RQ3 Do attention mechanisms provide meaningful explanations for which abilities and experiences drive matches?
• RQ4 What is the impact of incorporating historical recruitment data into learning the fit between job postings and resumes?

Key findings

• APJFNN outperforms all baselines on accuracy, precision, recall, F1, and AUC in the reported experiments.
• The full APJFNN model achieves the best metrics, with AUC of 0.8316 and accuracy 0.7559, surpassing BPJFNN.
• Attention mechanisms improve interpretability by highlighting crucial abilities and experiences.
• Incorporating word-level representations and hierarchical attention yields significant performance gains over traditional feature-based models.
• BPJFNN improves over non-attentive variants but still underperforms APJFNN, indicating the value of the proposed attention strategies.