메뉴 건너뛰기
.. 내서재 .. 알림
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)
로그인 회원가입 고객센터 ENG
주제분류

추천
검색

논문 기본 정보

자료유형
학위논문
저자정보

김천중 (고려대학교, 高麗大學校 大學院)

지도교수
임해창
발행연도
2016
저작권
고려대학교 논문은 저작권에 의해 보호받습니다.

이용수2

표지
AI에게 요청하기
추천
검색

이 논문의 연구 히스토리 (2)

초록· 키워드

오류제보하기
SNS(Social Network Service)의 발달로 자신이 좋아하는 유명인이나 친구의 게시글
을 읽고 댓글을 쓰는 사람이 많아졌다. 자기가 공감하는 게시글에는 호응하는 글을 자
기와 다른 생각을 가지고 있을 경우에는 비판이나 따끔한 충고를 하며 서로의 생각을
공유하는 장소가 되어야 하지만 몇 년 전부터 게시글과 전혀 관련이 없는 댓글들로 인
해 인터넷 문화가 역 방향으로 흘러가고 있다.
이러한 인터넷 문화를 바로 잡기위해 예전에는 특정단어가 포함되었는지 여부를 확
인하는 확률기반과 규칙기반 연구들이 진행되었고 현재는 SNS의 댓글에 관한 학습기
반 연구들이 진행되고 있다.
하지만 현재 진행되고 있는 연구들은 SNS의 댓글 자체만을 판단할 뿐 SNS에 등록
된 게시글과 댓글이 서로 연관성이 있는지를 확인하여 댓글의 적합성을 판단하는 연구
는 찾아 볼 수가 없었다.
본 연구에선 댓글 자체만을 판단하는 시스템이 아닌 SNS의 특수한 환경에서 나오
는 비정상적으로 짧은 글들과 비속어, 줄임말 등으로 단어 확률 기반 처리가 힘든 부분
을 고려하여 다양한 자질들을 발굴 하였고 이를 토대로 지지벡터머신의 자질로 사용하
여 댓글들이 해당 게시글과 적합한지를 판단하는 시스템을 구축하였다.
글 내용의 연관성과 관련된 Content 자질 군은 게시글과 댓글의 유사성 판단을 위한
11개의 세부 자질을 사용하였다.
감정 자질 군에서는 감정적으로 비슷한 단어의 감정 상태를 확인하기 위해 게시글과
댓글에 표출된 감정을 범주별로 구분하여 감정 단어 사전을 기반으로 게시글과 댓글의
감정 단어와 단어의 의미를 함축한 이모티콘 자질, 게시글을 긍정 또는 부정하는 댓글
자질로 구분 하였다.
Content 자질 군과 감정 자질 군은 게시글과 댓글의 내용과 관련되었다고 볼 수 있
다면 Surface 자질 군은 특수한 SNS 환경이 적합성 판단에 미치는 영향을 알아보기
위해 추가된 자질 군으로 게시자의 영향력과 시간에 따른 등록된 글들의 적합성을 판
단 할 수 있는 자질로 선정 하였다.
Contents 자질 군, 감정 자질 군, Surface 자질 군에 속한 총 18가지의 다양한 자질
들을 사용하여 분류 정확도를 향상시킬 수 있음을 보였고 자질 별 기여도를 확인하여
각 자질이 미치는 영향에 대해서 연구를 실시하였다.

목차

요 약 ·························································································································i
Abstract ··················································································································iii
목 차 ·······························································································································v
1 서 론 ······························································································································1
1.1 연구 배경 ····························································································································1
1.2 연구 목적······················································································································3
1.3 연구 구성····························································································································4
2 관련연구 ······················································································································6
2.1 국내외 사례···············································································································6
2.1.1 자질 관련 기존 연구·······················································································6
2.1.2 댓글 적합성 관련 기존 연구···························································································7
2.1.3 감정 분류 기존 연구·················································································8
2.2 기존연구의 문제점························································································9
2.3 기계 학습 기법 및 자질 선정 기법·························································································9
2.3.1 지지 벡터 머신 (Support Vector Machine)······························································10
2.3.2 정보 이득(Information Gain)··········································································12
3 적합성 판단 설계 및 구현 ·····························································································14
3.1 적합성 판단 전체 시스템 구성도·························································································14
3.2 전처리 과정·······································································································15
3.3 댓글 적합성 판단 자질·····························································································17
3.3.1 Content 자질··································································································18
3.3.2 Surface 자질····································································································22
3.3.3 감정 자질····································································································24
4 실험 및 결과 ·········································································································28
4.1 실험 환경··········································································································28
4.1.1 실험 데이터·····························································································28
4.1.2 평가 척도·········································································································30
4.1.3 정보이득 실험 구축··························································································30
4.2 실험 결과··············································································································32
4.2.1 모든 자질 군 결과 확인···························································································32
4.2.2 자질 군 결과 확인························································································33
4.2.3 자질 결과 확인····························································································33
4.2.4 자질 기여도 수치 확인······························································································34
4.3 실험 결과 분석··································································································36
4.3.1 적합성 판단 시스템 및 자질들의 유효성·····································································36
4.3.2 자질 기여도 분석··································································································36
5 결론 및 향후 과제 ····································································································38
참고문헌 ·························································································39
감사의 글 ···························································································41

최근 본 자료

전체보기

댓글(0)

0