CV Content Recognition and Organization Framework based onYOLOv8 and Tesseract-OCR Deep Learning Models

Abstract The growing number and variety of resumes in the job market makes it necessary to have better systems for sorting them efficiently. The task of finding suitable candidates for an open job position would be a repetitive and time-consuming task, especially from a large pool of candidates. The same task could be an actual fair screening and shortlisting though. Indeed, it is not acceptable to lose the chance to employ top skilled candidates because of the tardiness of the whole census or the poor selection caused by human fault. In order to help automating this cycle, this study introduces a new CV recognition system that combines advanced technologies: You Only Look Once (YOLO) for detecting important sections in the CV files, Tesseract-OCR for extracting text in each section, and an efficient post processing steps for correcting possible faults in recognized text. We also propose to automatically organize the data within the CV into a database to facilitate performing any data analytics or searching processes. To evaluate the system, we used a dataset of 1,300 resumes in JPEG, PNG, and JPG formats from various sources, showcasing different formats, languages, and quality levels. Preprocessing steps were taken to ensure the data is high quality and consistent. This approach saves time and improves the efficiency and accuracy of sorting resumes, helping HR teams focus on more strategic tasks and making the hiring process easier and less stressful. Results support the validity of the proposed system through the experiments with this diverse dataset. The system’s effectiveness was confirmed through experiments with this diverse dataset, achieving a mean Average Precision (mAP) of 92.1%, a precision rate of 92.2%, and a recall rate of 86.0%.