CareerSmart: An Intelligent Recruitment Framework Using TF-IDF, BERT, and Named Entity Recognition for Resume-Job Matching

This paper introduces CareerSmart, a smart recruitment system that automatically uses rule-based and semantic natural language processing algorithms to screen resumes and rank the shortlisted candidates in a deployable web-based system. The suggested framework helps to ingest multi-format resumes (PDF, DOCX/DOC, TXT/RTF/ODT and image files) and use OCR fallback when scanning documents, and then text normalization with further feature extraction is suggested. The calculation of the candidate-job alignment includes a hybrid scoring pipeline that counts dynamically lexical coverage of skills by the keyword at the same time, lexical relevance TF-IDF cosine similarity, and semantic similarity scores based on the transformer-based Sentence-BERT embeddings, and adaptive weighting preserves the stability of the results under the unavailability of profound semantic components. The system is also more interpretable as it features named entity recognition and automatic feedback generation, finding matching competencies, those that are not present, and constructive improvements recommendations that applicants can take. The platform, developed on Flask and SQLite, implements end-to-end workflows of hiring, such as job posting, applicant applications, resume auto alcoholism, ranking, shortlist and rejection notification, and can be implemented in small to medium hiring companies. The application in real working conditions demonstrates that the hybrid methodology can be better at matching robustness than either of the single-method baselines, as it involves both explicit skill references and context-based semantic markers, lessening the effort of human recruiters and giving more open, evidence-based guidance to fairer and more efficient hiring results.