Human resource management is a strategic axis for organizations, especially in contexts where artificial intelligence (AI) tools, such as natural language processing (NLP), play a fundamental role. Recruiting external applicants from large CV repositories requires consistent screening. The proposed methodology involves leveraging an existing curriculum vitae (CV) repository, structuring and indexing the data within a vector-based knowledge base, and applying retrieval techniques to identify candidates that satisfy role-specific criteria. Using 5029 CVs as benchmarks, we evaluate 3 queries, 3 variables (Degree, Skills, Experience), and 7 scenarios. Sampling n = 76 CVs for Queries 1–2 and n = 350 CVs for Query 3. The proposed approach achieved consistently high specificity across scenarios and query profiles, while sensitivity showed the largest fluctuations, particularly under single-requirement configurations. Across all queries and scenarios, accuracy ranged 65.79–98.00%, specificity 86.67–100.00%, and sensitivity 0.00–94.92%, while error rates decreased from 34.21% to 2.00% as constraint strictness increased. Sensitivity fluctuated most under single-requirement settings, and Experience-only screening showed the weakest selection behavior. Moreover, the results indicate that the ability to confirm suitable candidates is sensitive to query formulation, since non-standard role naming, experience phrasing, and other lexical variations can reduce the system’s capacity to detect positive evidence. Overall, these findings indicate that a knowledge-base-centered design enables consistent and interpretable requirement-driven candidate screening and provides a quantitative baseline for future improvements in recruitment-oriented retrieval systems.
Morales-Zaleta et al. (Wed,) studied this question.