DeepSkillNER: An automatic screening and ranking of resumes using hybrid deep learning and enhanced spectral clustering approach

The process of identifying the best job candidates from different sets of resumes is a resource and time consuming process. The common issue in resume screening is the unavailability of annotated data to label the information obtained accurately. To tackle these challenges, this research undergoes feature extraction and feature clustering stages for resume screening and ranking. In this work, the hybrid deep learning (DL) based Pyramid Dilated Convolutional Neural Network with Bidirectional Gated Recurrent Unit (PDCNN-Bi-GRU) is introduced to extract the skill related features from the resumes. Moreover, the Conditional Random Field (CRF) layer is hybridized with the DL technique to generate skill-related information as the outcome. Then, a fuzzy matching module is contemplated to match the skill-related features with the job categories to enhance the accuracy performance. Finally, the candidate resumes of higher skills are clustered using the hybrid Spectral clustering with Hummingbird Optimization (SCHO) technique. The proposed method is tested with both public source and real-time datasets and is implemented in the PYTHON platform. The performances like accuracy, specificity, sensitivity, F-measure, kappa, time complexity, and Mean Square Error (MSE) are analyzed and compared with existing techniques. The proposed method obtains the accuracy of 99.3% and 99.83% for testing with both public source and real-time datasets, respectively.