Recruitment systems play a vital role in ensuring that qualified candidates are matched with organizational needs. Yet traditional approaches, such as keyword-based searches and manual filtering, often fail to capture the semantic context of resumes and job descriptions. This study presents a systematic literature review on the use of natural language processing (NLP) and deep learning models in automated recruitment and human resource information systems. Specifically, the study addresses two research questions: (1) What NLP techniques are used in modern HR recruitment systems? and (2) How effective are BERT and other deep learning models in automating recruitment tasks such as resume screening and job-candidate matching? The review reveals a progression from rule-based and statistical models, such as TF-IDF, toward embedding-based methods like Word2Vec and GloVe, culminating in transformer-based models, including BERT. Findings indicate that BERT and its variants consistently outperform earlier methods in precision, recall, and F1-score, achieving superior accuracy in parsing candidate qualifications and assessing semantic similarity. However, challenges remain regarding computational demands, scalability, and organizational adoption. The study contributes by synthesizing current approaches, highlighting performance benchmarks, and offering recommendations for developing recruitment systems tailored to higher education institutions.