지원사업
학술연구/단체지원/교육 등 연구자 활동을 지속하도록 DBpia가 지원하고 있어요.
커뮤니티
연구자들이 자신의 연구와 전문성을 널리 알리고, 새로운 협력의 기회를 만들 수 있는 네트워킹 공간이에요.
논문 기본 정보
- 자료유형
- 학위논문
- 저자정보
- 지도교수
- 권혁철
- 발행연도
- 2021
- 저작권
- 부산대학교 논문은 저작권에 의해 보호받습니다.
이용수7
이 논문의 연구 히스토리 (2)
- 이 논문의 후속연구가 궁금하신가요?
- 연관 학술논문 또는 학술발표를 통해 보다 발전된 연구결과를 확인하실 수 있습니다.
- 이 논문의 연구 히스토리 확인하기
초록· 키워드
오류제보하기
본 연구에서는 질의응답에 다수 사용되는 한국어의 수사를 포함한 질의응답을 처리하기 위해 대한민국 법령문서를 이용하여 관련 데이터 세트를 구축하였다. 이를 BERT 기반 전이학습을 이용한 언어모형에 훈련하여 한국어 수사를 포함한 문서를 위한 질의응답 시스템을 개발했다. 한국어 수사를 처리하기 위해 한국어의 수사, 수분류사, 분류사, 물음말을 이용하고 일반화를 위해 국제단위계를 이용하여 데이터 세트를 구축했다. 또한, 법령문서의 구조적 특징을 반영한 문서 가공 방식을 제안했다. 데이터 세트의 일반화를 알아보기 위해 한국어 수사를 포함한 문서를 위한 질의응답 데이터 세트와 전자 신문 기사를 이용한 평가 데이터 세트로 실험을 수행했다. 실험 결과는 구축한 시스템이 EM 82.91, F1 90.95로 기존과 비교하여 EM 13.57, F1 9.55만큼 향상된 성능을 보였다. 생성한 데이터 세트 평가를 위해 한국어 대용량 질의응답 데이터 세트인 KorQuAD의 검증 데이터 세트 중 한국어 수사를 포함하는 질의응답을 추출하여 실험을 수행했고, 이 또한 EM 13.12, F1 7.15만큼 향상된 성능을 보였다.
목차
I. 서론 ·································································································································· 11. 연구 배경 및 필요성 ············································································································ 12. 연구 목적 및 범위 ················································································································ 23. 연구 방법 및 절차 ················································································································ 3II. 관련 연구 ······················································································································· 51. 질의응답 데이터 ···················································································································· 52. 사전훈련 언어모형 ················································································································ 83. 한국어 수사(數詞) 처리 ····································································································· 144. 법령문서를 활용한 연구 ···································································································· 17III. 시스템 및 데이터 세트 구축 ················································································· 201. 한국어 수사를 포함한 문서를 위한 질의응답 시스템 구축········································· 202. 데이터 세트 구축 ················································································································ 233. 기계 학습 및 미세조정 ······································································································ 314. 평가 방법 ······························································································································ 345. 실험 결과 및 검증 ·············································································································· 37IV. 결론 ····························································································································· 41참고문헌 ···························································································································· 43
최근 본 자료
댓글(0)
0