Job searching can be complex and time-consuming, requiring candidates to sift through numerous job offers to find those that best match their skills and aspirations. Recommendation systems are crucial in streamlining this process by suggesting relevant job opportunities based on user profiles. This research addresses the challenge of improving job recommendations by leveraging the semantic information contained in job documents, such as descriptions and resumes. We propose a hybrid approach that combines BERT, a language model proficient in understanding text context, with WordNet, a lexical database rich in semantic relationships. This integration aims to enhance the relevance of recommendations by capturing detailed semantic nuances. Evaluation results indicate that incorporating WordNet with BERT generally improves recommendation performance by 10%. However, the effectiveness varies across different profiles, with some cases experiencing reduced relevance due to semantic noise introduced by WordNet. Despite this, the approach demonstrates significant potential for enhancing job recommendations using advanced natural language processing techniques.