TOWARDS BUILDING A HYBRID APPROACH FOR CONCEPTUAL-BASED CLASSIFICATION AND RANKING OF RESUMES AND THEIR CORRESPONDING JOB POSTS رسالة ماجستير

Abstract

With the increasing growth in online recruitment, traditional hiring methods are becoming inefficient. This is due to the fact that job portals receive enormous numbers of unstructured resumes - in diverse styles and formats - from applicants with different fields of expertise and specializations. Therefore, the extraction of structured information from applicant resumes is needed not only to support the automatic screening of candidates, but also to efficiently route them towards their corresponding occupational categories. This process assists in minimizing the effort required by employers to manage and organize resumes, as well as to screen out irrelevant candidates. Several techniques and approaches have been proposed to address the issue of automatic matching between resumes and job postings. However, little attention has been paid to address problems associated with classification of resumes and job posts, automatic ranking of applicant resumes, and automatic profile generation from applicants’ resumes. In this research work, we present a Job Post and Resume Classification system that exploits an integrated knowledge base for carrying out the classification task. Unlike conventional systems that attempt to search globally in the entire space of resumes and job posts, the proposed approach matches resumes that only fall under their relevant occupational categories. In addition, our proposed system attempts to exploit the extracted information from applicants’ resumes to automatically generate user profiles that can be further used for recommending jobs to job seekers. In this context, our proposed approach attempts to push job post notifications that satisfy job seekers’ preferences and skills. To demonstrate the effectiveness of the proposed system, we have conducted several experiments using a real-world recruitment dataset. Additionally, we have evaluated the efficiency and vi effectiveness of proposed system against state-of-the-art online recruitment systems and the results were published in two well-recognized international conferences in 2017