在全球化與數位轉型的競爭環境下，大型企業每年面對數以萬計的應徵履歷，傳統人工書面審查耗時費力，也容易因審閱者疲勞與主觀偏差影響甄選的信度與效度。儘管人工智慧已逐步導入招募流程，現有招募系統仍存在許多限制，包括難以同時整合應徵者的顯性資歷與自傳中所傳達的隱性特質、對中文自傳語意的判別能力有限，以及缺乏對企業內部甄選規則的整合，使模型建議的結果未必符合組織的實務決策邏輯。 為了解決上述問題，本研究提出整合結構化資歷、自傳文本與企業內部關鍵條件之招募篩選模型「對比式自動化招募篩選」（Contrastive Automated Recruitment Screening， CARS）。CARS 同時處理應徵者的背景資料與自傳內容，並將企業內部的甄選規則獨立編碼，使模型於判別時能在語意層面理解應徵者特質的同時，仍保留對實務硬性條件的敏感度。 為驗證CARS模型的實用性，本研究以某大型企業真實招募履歷進行實驗。實驗結果顯示，CARS於準確度、精確度、召回率與F1分數四項指標皆優於傳統的人工智慧篩選模型，消融實驗也證實三項核心設計皆對整體效能有實質貢獻。個案研究進一步驗證CARS於實務情境下的判別能力，模型對於通過與未通過樣本皆能做出與實際甄選結果一致的預測，並能有效識別語意上看似匹配但不符合企業內部甄選條件的應徵者。綜合上述結果，CARS可作為企業組織在大規模招募中協助書面審查階段的可行工具。;In large-scale recruitment， manual resume screening is time-consuming and can be affected by reviewer fatigue and personal bias. Current AI-based screening tools still have several problems. They often struggle to combine a candidate′s background information with the personal traits shown in their autobiographical text. They also have trouble understanding Chinese self-descriptions written by job applicants， and they usually do not include the internal screening rules that companies use， which makes the model′s suggestions different from real hiring decisions. To solve these problems， this study proposes Contrastive Automated Recruitment Screening (CARS)， a recruitment screening model that combines structured background data， autobiographical text， and company-specific screening rules. CARS looks at both a candidate′s background and self-description together and handles the company′s screening rules separately. This lets the model understand what a candidate is like， while still checking whether they meet the company′s basic requirements. Experiments using real recruitment data from a large company show that CARS performs better than common AI screening models in accuracy， precision， recall， and F1-score. Ablation studies show that all three main design parts contribute to the model′s performance. Case studies also show that CARS makes predictions that match actual hiring results and can spot candidates who look like a good fit on paper but do not meet the company′s requirements. Overall， CARS can serve as a useful tool for companies to support the resume screening process in large-scale recruitment.
