Optimizing Recruitment with BERT: A Multi-task Approach for Resume Screening and Candidate Ranking

This paper suggests a BERT-based framework of multi-task learning to realize the intelligent movement of resume screening and ranking of candidates and employs explainable AI (XAI) techniques to provide transparency to the model. A BERT architecture is a pre-trained model that captures contextualized embeddings to extract resumes and thus classify and rank them properly. Experimental findings prove it works better, and the model shows 92 percent levels of accuracy in resume screening or 88 percent Mean Average Precision (MAP) in candidate ranking. These findings are much better compared to the conventional machine learning model, including SVM and Random Forest, which attained 85%-86% classifier and 75%-77% MAP, in resume screening and candidate ranking, respectively. SHAP and LIME integration guarantee transparency, and such features as skills, experience are highlighted as the most influential in decision-making. Further, it is confirmed that model performance with the AUC of 0.96 in the ROC curve is strong with regards to discrimination capacity by utilizing a confusion matrix. Loss and accuracy over epochs plots show that the generalization metrics of the model were steadily improving along with training and validation accuracy. In addition to enhancing the manual process of screening resumes and ranking the applicants, the proposed framework offers insight into interpretation of the actual process of decision-making to be used by the hiring company in the recruitment process. The paper illustrates the possibility of employing BERT-based multi-task learning models, alongside explainable AI, to improve the recruiting process and offer a clear, effective, and scalable analysis of analyzing candidates via the use of an automated candidate assessment mechanism.

• Publication year

  2025

• Venue

  2025 2nd Global AI Summit - International Conference on Artificial Intelligence and Emerging Technology (AI Summit)

• Publication date

  2025-11-19

• Fields of study

  Not labeled

• Identifiers
  DOI 10.1109/AISummit66170.2025.11410770
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Spotlighting recruitments: is AI dominating human resource practices? Qualitative research using NVIVO

  2025cited by this paper
• GSR‐C2N: Graph Feature Extracted Spar‐Raven Optimized CNN Based Crypto Mining Framework

  2025cited by this paper
• MetaV: A Pioneer in feature Augmented Meta-Learning Based Vision Transformer for Medical Image Classification

  2024cited by this paper
• Navigating Water Scarcity with IoT: A Smart Management System Approach

  2024cited by this paper
• Optimizing Water Resources: An IoT Smart Management System Analysis

  2024cited by this paper
• Innovative Waste Management: The Efficiency of IoT-Enabled Smart Bins

  2024cited by this paper
• Candidate Evaluation with Multimodal Data-Driven for Recruitment

  2024cited by this paper
• Semantic Analysis Using Deep Learning for Predicting Stock Trends

  2023cited by this paper
• A Comprehensive Study on Video Captioning Techniques, Benchmark Datasets and QoS Metrics

  2023cited by this paper
• Evaluating LSTM and GRU Models for Cryptocurrency Price Forecasting in Financial Markets

  2023cited by this paper
• Emotion Analysis for predicting the emotion labels using Machine Learning approaches

  2021cited by this paper
• Building an ASR System for Indian (Punjabi) language and its evaluation for Malwa and Majha dialect: Preliminary Results

  2021cited by this paper
• Grading and sorting technique of dragon fruits using machine learning algorithms

  2021cited by this paper
• A Review on Multisource Data Analysis using soft computing Techniques

  2018cited by this paper

• No citing papers are available for this paper.