Large Language Model-based Resume-Job Intelligent Matching Algorithm and Its Adaptability Case Study Across Different Positions in the Hotel Industry

As a service-intensive industry, the hotel sector exhibits significant variations in talent requirements across different positions, making traditional resume screening methods inadequate for meeting the complex matching demands of various roles. This study develops a resume-job intelligent matching system specifically designed for the hotel industry based on large language model technology and conducts an in-depth analysis of its adaptability performance across different position types. Using a chain hotel group as a case study, we collected 1,847 historical recruitment records spanning nearly 10 years (2014-2024) across six major categories: front office, housekeeping, food & beverage, sales, logistics, and management. Through controlled experiments and ablation studies, we systematically evaluated the performance differences of large language models in resume screening across various position types. The experiments employed LoRA fine-tuning techniques for domain adaptation of the BERT model and designed a multi-dimensional matching algorithm integrating skills, experience, and soft skills evaluation dimensions. Results demonstrate that the system achieves significantly higher matching accuracy in highly standardized positions (front desk reception: 89.2%, housekeeping: 87.6%) compared to positions requiring strong personalization (sales manager: 76.8%, F&B supervisor: 78.1%). Compared to traditional TF-IDF methods, the F1 score improved by 32.1 percentage points with statistical significance (p < 0.001).