AI Resume Screening System
Moneeb Raza
Department of Computer Science and Engineering Galgotias University, Greater Noida, India muneebraza167@gmail.com
Md Sharique
Department of Computer Science and Engineering Galgotias University, Greater Noida, India mdsharique8757@gmail.com
Abstract—AI-based resume screening systems are widespread and represent online web applications placed at the stage preced-ing applicant tracking systems. The former automatically gather information from candidates’ documents, evaluate candidates’ profiles on the basis of requirements, score their suitability and recommend candidates for further hiring processes. In this research, resume screening is considered as a task involving web systems and machine learning techniques. We outline the whole pipeline of building browser-based resume processing that involves NLP and ML-based extraction of data, server-side normalization and matching, and a graphical user interface that shows the scores with corresponding feedback. Although TF-IDF (cosine similarity) is widely used as a foundation for many practical applications and research studies, significant problems associated with its lack of interpretability, transparency and fairness, as discussed in previous literature, have been identified. Most existing systems lack explanations for decisions and often show weak ability to detect bias with limited feedback for applicants and recruiters.
In order to fill the mentioned gap, a web-first architecture combining resume parsing and section segmentation with a hybrid scoring technique based on both lexical and semantic features was proposed. Such solution incorporates explainability module, periodic checking for fairness, preprocessing considering privacy issues, and a human override logging and customization of weights and filtering in dashboards. The metrics considered in the evaluation include those reflecting the ranking performance and system capabilities, as well as the criteria specific for deployment like latency and usability. Our results, obtained from the analysis of the relevant literature and implementation process, suggest that proposed systems can shorten the screening time, select better qualified candidates and make hiring decisions more transparent. However, the success and sustainability of these solutions depend on the implementation of appropriate gover-nance, fairness and privacy requirements during development. Index Terms—Resume Screening, Natural Language Processing, TF-IDF, Cosine Similarity, Applicant Tracking System, Fairness, Explainability.
Index Terms—Resume Screening, Natural Language Process-ing, TF-IDF, Cosine Similarity, Applicant Tracking System, Fairness, Explainability
