CAJ | 학술논문

文本倾向性分析已成为当前自然语言处理领域的研究热点，其研究成果具有极高的应用价值。针对网络在线中文评论的特点，基于领域本体与情感词典对商品评论倾向性进行分析。其主要思想是首先构建面向商品论坛的领域本体；其次利用情感词典与上下文极性算法计算情感词极性；再次通过将本体与SBV算法相结合，实现评价对象和评价词的二元组抽取；最后完成句子的倾向性分析。实验结果表明，有效提高了句子级倾向性分析的准确率。
문본경향성분석이성위당전자연어언처리영역적연구열점，기연구성과구유겁고적응용개치。침대망락재선중문평론적특점，기우영역본체여정감사전대상품평론경향성진행분석。기주요사상시수선구건면향상품론단적영역본체；기차이용정감사전여상하문겁성산법계산정감사겁성；재차통과장본체여SBV산법상결합，실현평개대상화평개사적이원조추취；최후완성구자적경향성분석。실험결과표명，유효제고료구자급경향성분석적준학솔。
Text propensity analysis has currently become the focus of research in natural language processing field, its research results have the extremely high application value.Aiming at the characteristics of online internet Chinese reviews, in this paper we analyse the tendencies of product reviews based on domain ontology and sentiment lexicon.Our main idea is to build the product forum-oriented domain ontology first. Then we calculate the polarities of sentiment words by using the sentiment lexicon and context polarity algorithm.Thirdly, by combining domain ontology with SBV algorithm we realise the extraction of 2-tuple of evaluation object and evaluation words.Finally, we complete the analysis of sentence propensity.Experimental results show that it improves the accuracy of propensity analysis in sentence level effectively.