计算机技术与发展
計算機技術與髮展
계산궤기술여발전
COMPUTER TECHNOLOGY AND DEVELOPMENT
2015年
5期
104-109
,共6页
HowNet%情感词汇本体%同义词词林%等价情感倾向集合
HowNet%情感詞彙本體%同義詞詞林%等價情感傾嚮集閤
HowNet%정감사회본체%동의사사림%등개정감경향집합
HowNet%affective lexicon ontology%Tongyici-Cilin%emotion-equal set
文本情感倾向性判别是情感分析的重要组成部分，而精确的词汇情感计算是文本情感倾向性判别的基础。基于词汇知识库的情感词倾向判别计算引起了学者们的广泛关注与研究。文中融合国内知名的三大词典：HowNet、同义词词林、情感词汇本体，重新对基准词对做进一步的归纳与总结，从不同的角度构建三类等价情感倾向集合并提出两种处理集合的策略，建立了待定情感词与特定等价情感倾向集合的情感映射关系。实验结果表明：该方法获得的最高准确率可达91.62%，平均正确率85.31%，符合人们对词语的情感倾向认识。
文本情感傾嚮性判彆是情感分析的重要組成部分，而精確的詞彙情感計算是文本情感傾嚮性判彆的基礎。基于詞彙知識庫的情感詞傾嚮判彆計算引起瞭學者們的廣汎關註與研究。文中融閤國內知名的三大詞典：HowNet、同義詞詞林、情感詞彙本體，重新對基準詞對做進一步的歸納與總結，從不同的角度構建三類等價情感傾嚮集閤併提齣兩種處理集閤的策略，建立瞭待定情感詞與特定等價情感傾嚮集閤的情感映射關繫。實驗結果錶明：該方法穫得的最高準確率可達91.62%，平均正確率85.31%，符閤人們對詞語的情感傾嚮認識。
문본정감경향성판별시정감분석적중요조성부분，이정학적사회정감계산시문본정감경향성판별적기출。기우사회지식고적정감사경향판별계산인기료학자문적엄범관주여연구。문중융합국내지명적삼대사전：HowNet、동의사사림、정감사회본체，중신대기준사대주진일보적귀납여총결，종불동적각도구건삼류등개정감경향집합병제출량충처리집합적책략，건립료대정정감사여특정등개정감경향집합적정감영사관계。실험결과표명：해방법획득적최고준학솔가체91.62%，평균정학솔85.31%，부합인문대사어적정감경향인식。
Text sentiment orientation analysis is an important part of textual emotion classification,and to determine the emotional bias for words precisely is the very foundation of text sentiment orientation analysis. Nowadays,dictionary-based methods of sentiment orientation computing for words have been widely concerned and researched. In this paper,summarize and conclude benchmark words considering knowledge from HowNet,Tongyici-Cilin and affective lexicon ontology,then build three emotion-equal sets from different perspective and propose two strategies of set processing,finally establish the mapping between the sentiment orientation of a word to be computed and a specific emotion-equal set. Experimental results has generally agreed with the sentiment orientation felt by human,and achieved maxi-mum correctness 91. 62% and 85. 31% in average.
