A New Approach on AI-Based Resume Analysis and Feedback System

Abstract

Traditional Applicant Tracking Systems (ATS) are based mostly on filtering based on keywords and usually do not provide the semantic meaning, structural quality, and contextual relevance of resumes. Such a restriction causes incorrect analysis of candidates and lower recruitment transparency. In order to deal with these issues, this paper suggests a resume analysis and feedback system based on AI which incorporates all three concepts of semantic similarity, relevance, and heuristic structural analysis into an integrated hybrid framework. This suggested system presents a Hybrid Semantic–Keyboard–Heuristic (HSK) scoring split, a hybrid of rule-based and contextual evaluation to offer balanced resume scores. The overall correspondence between resumes and job descriptions is captured with the help of semantic similarity, and the necessary skills are identified with the help of key words matching. The heuristic aspect considers the resume format, completeness and readability. This is an operation that gives a weighted average of these elements to get the final score. Moreover, it will use a Gemini-based Large Language Model to provide customized section-by-section feedbacks that allow candidates to effectively amend their resumes. The results of the experiment show better accuracy, interpretability and correspondence with human assessment than traditional ATS methods. The suggested system is a clear, understandable, and easy-to-use solution to resume evaluation in the current era.

Introduction

The text describes an AI-powered resume evaluation system designed to improve traditional Applicant Tracking Systems (ATS), which mainly rely on keyword matching and often unfairly reject qualified candidates due to formatting or wording differences.

The proposed system uses Natural Language Processing and Large Language Models to understand resumes more deeply by combining three evaluation components: semantic similarity (meaning-based matching), keyword relevance (important skills from job descriptions), and heuristic scoring (resume structure and quality). These are combined into a hybrid scoring model that ranks candidates more fairly and accurately than traditional systems.

A key feature is section-wise resume analysis (education, skills, experience, etc.) and LLM-generated personalized feedback to help applicants improve their resumes. The system also visualizes results using charts for better interpretability.

Existing approaches—such as TF-IDF, machine learning classifiers, embeddings (like SBERT), and GPT-based tools—are reviewed, highlighting limitations like lack of context understanding, no structured scoring, or absence of personalized feedback. The proposed method addresses these gaps by integrating semantic understanding, structured scoring, and explainable AI feedback into a single framework.

Conclusion

The paper introduced a new AI-powered resume analysis and feedback system, which combines semantic similarity, key-word relevance, and heuristic structural analysis into a hybrid scoring system. The suggested model of HSK offers a clear and consistent way of evaluating resumes, which is focused on the shortcomings of traditional ATS systems based on the use of keywords. The system enhances the precision and unbiasedness of the evaluation of candidates by factoring in the contextual knowledge and systematic examination. The system is further augmented with the integration of a Gemini-based Large Language Model to give personalized and section-wise feedback so that users can effectively improve their resumes. Results interpretability is enhanced by the use of visualizations (radar and bar charts) which makes the system easier to use and more informative. Experimental findings indicate that the suggested method is in line with the human assessment and has better performance than the traditional techniques. Besides screening the resumes, the system is an excellent guide to candidates and is of great help to job seekers and employers. The system can be expanded in the future to help in multilingual resume analysis, domain specific scoring models and real time resume optimization. The system can also be made more robust by incorporating machine learning methods of adaptive weighting and bias detection. Through these enhancements, the proposed framework has great potential to become a developed, industry-savvy ATS solution.

References

[1] Devlin, M. Chang, K. Lee, and K. Toutanova, “BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding,” Proc. NAACL, 2019. [2] N. Reimers and I. Gurevych, “Sentence-BERT: Sentence Embeddings using Siamese BERT Networks,” EMNLP, 2019. [3] A. Vaswani et al., “Attention is All You Need,” Advances in Neural Information Processing Systems, 2017. [4] M. Gupta and P. Gupta, “Automated Resume Screening Using NLP and Machine Learning,” International Journal of Advanced Research in Computer Science, vol. 10, no. 5, 2019. [5] A. Kumar and R. Garg, “Semantic Similarity Based Resume Ranking Using Deep Learning,” International Conference on Intelligent Computing and Control Systems (ICICCS), 2021. [6] S. Sarkar and P. B. Mandal, “A Hybrid Approach for Resume Classification Using Machine Learning and Text Mining Techniques,” IEEE Access, vol. 8, pp. 136348–136360, 2020. [7] Google DeepMind, “Gemini: A Family of Highly Capable Multimodal Models,” Technical Report, 2024. [8] J. Kowsari et al., “Text Classification Algorithms: A Survey,” Information, vol. 10, no. 4, 2019. [9] H. Chen, J. Fan, and Z. Wang, “Automated Job Matching and Recommendation Using NLP Techniques,” IEEE Transactions on Computational Social Systems, vol. 7, no. 4, 2020. [10] S. A. Raza and M. A. Khan, “Evaluating Applicant Tracking Systems: Limitations and Future Directions,” International Journal of Computer Applications, vol. 182, no. 43, 2019.

Copyright

Copyright © 2026 Priyanshu Verma, Isha Choudhary , Janhvi Shankar Singh , Kumari Nency, Sambit Satpathy , Dhirendra Kumar Shukla. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.