International Research Journal of Education and Technology

This study introduces a resume analyzer using large language models (LLMs) to automate and enhance the classification of resumes in hiring databases. Traditional resume classification, typically done manually, is time consuming and prone to inconsistency, especially as online recruitment continues to grow. Conventional machine learning approaches to resume categorization face limitations due to sparse labeled training data, scalability issues, and variable data quality. Our proposed method leverages a graph multi headed attention network (MGAT) model in a domain adaptation framework, trained on structured job post data to classify unstructured resume data. By treating the job post dataset as the source domain and the resume dataset as the target domain, this approach enhances classification accuracy and reduces reliance on extensive resume training data. The MGAT-based solution also addresses the challenges posed by long, variably formatted resumes, improving both classification efficiency and semantic alignment between resumes and job postings. This work represents a pioneering application of graph neural networks in the domain adaptation context for resume classification, presenting a scalable and effective solution for recruitment automation.

Key Words— Resume Classification, Large Language Models (LLMs), Recruitment Automation, Semantic matching, Machine Learning, Graph Multi-Headed Attention Network (MGAT)