Developing an Intelligent Job Recommendation System Using Semantic Retrieval and Explainable AI Techniques

Combining different search methods improves job matching: The system uses both traditional keyword matching (TF-IDF) and modern semantic understanding (Sentence-BERT) together, achieving strong performance metrics (80.32% precision) by capturing both exact matches and jobs with similar meanings expressed in different words.

Explainability is built into the recommendation process: Unlike many AI systems that act as "black boxes," this approach generates explanations for why specific jobs are recommended, making it easier for job seekers to understand and trust the system's suggestions.

Privacy-conscious design protects user data: The system relies only on structured job metadata (title, location, company, seniority level) rather than tracking user browsing history or personal interaction data, reducing privacy concerns while still delivering effective recommendations.

Semantic understanding solves real recruitment problems: The research demonstrates that keyword-based search alone misses relevant opportunities when employers use different terminology for equivalent roles—a common issue that semantic retrieval methods can address, potentially helping job seekers find suitable positions they might otherwise overlook.

Strong performance on real-world data: Testing on over 31,000 actual LinkedIn job postings shows the system works effectively at scale, with the best configuration achieving an nDCG score of 0.9496, indicating high-quality ranking of recommended positions.