Hybrid Transformer-Based Resume Parsing and Job Matching Using TextRank, SBERT, and DeBERTa

As the job markets have become very competitive, distinguishing oneself has become very important. Resume summarization and ranking have emerged as a crucial task for processing and handling large volumes of resumes. With the advent of Natural Language Processing (NLP) techniques, automated resume summarization has gained significant attention due to its potential to expedite the hiring process while ensuring fairness and objectivity. This paper focuses on hybrid transformer-based resume parsing and job matching by implementing TextRank, SBERT, and DeBERTa. DeBERTa, an advanced transformer, functions on a disentangled attention mechanism for contextual understanding of the words, TextRank, SBERT and PageRank algorithms for extractive summarization. The ranking of candidates is done by calculating a composite score, which includes evaluation metrics like cosine similarity for job description match based on understanding the context. Difflib- a sequence matcher for candidates' experience fit, and Jaccard similarity for skills match. These scores are weighted based on their importance to the job, creating a balanced and tailored ranking. This approach focuses on saving time, reducing labour costs, and making recruitment more efficient by identifying the best matches for each position.