메뉴 건너뛰기
소속 기관 / 학교 인증
인증하면 논문, 학술자료 등을  무료로 열람할 수 있어요.
한국대학교, 누리자동차, 시립도서관 등 나의 기관을 확인해보세요
(국내 대학 90% 이상 구독 중)
고객센터 ENG
주제분류

논문 기본 정보

저자정보
(Handong Global Univ) (Handong Global Univ.) (Handong Global Univ.) (Kyungpook National Univ.)
저널정보
국제인공지능학회 The International Journal of Internet, Broadcasting and Communication The International Journal of Internet, Broadcasting and Communication Vol.18 No.5
오류 신고하기

피인용 0

검색

    초록·키워드

    Occupation–skill mapping requires retrieval methods that capture semantic similarity while respecting the structured relations between occupations and skills. We evaluate the retrieval component of a HybridRAG prototype in which we add a graph validation step to vector retrieval, retaining only candidates that remain directly connected in a knowledge graph derived from the Singapore Qualification Framework. We compare vector-only retrieval with vector retrieval followed by graph validation over 87,664 ground-truth query–target pairs. We find that graph validation improves Precision@1 from 0.739 to 0.928 and MRR from 0.804 to 0.928, while increasing average latency from 174.423 ms to 180.871 ms. We also show that the validator discards non-relevant candidates with 99.99% precision but returns no candidate for 7.2% of queries, which bounds the attainable Precision@1. Our contribution is a structure-preserving filter that suppresses semantically plausible but structurally invalid matches in a HybridRAG workforce-mapping pipeline at a small runtime cost.

    최근 본 자료 전체보기