CAJ | 학술논문

网络舆情参与者对舆情事件情感指向的清晰定位，可以为舆情事件主体或监管部门准确掌握民众关注焦点、制定有效的网络舆情引导措施提供科学的参考。使用PLSA模型提取网络舆情话题及其特征词，结合TF-IDF修正话题特征词；基于已经提取到的话题文本特征词，构建对应的情感词表，并应用HowNet相似度算法计算每个情感词对应的正负情感倾向值；综合考虑特征词对应的所有情感词，计算得出特征词的情感值，对舆情参与者的情感指向及变化进行准确定位。
망락여정삼여자대여정사건정감지향적청석정위，가이위여정사건주체혹감관부문준학장악민음관주초점、제정유효적망락여정인도조시제공과학적삼고。사용PLSA모형제취망락여정화제급기특정사，결합TF-IDF수정화제특정사；기우이경제취도적화제문본특정사，구건대응적정감사표，병응용HowNet상사도산법계산매개정감사대응적정부정감경향치；종합고필특정사대응적소유정감사，계산득출특정사적정감치，대여정삼여자적정감지향급변화진행준학정위。
The research on the evolution of netizens' sentiments provided theoretical basis for relevant organizations to understand the neti-zens' sentiments of the topics they were focusing on and to make effective opinion guidance of the network public topics. This paper used the PLSA model to extract the topics and the text-words matrixes after segmenting the texts. Combining with the TF-IDF model and the PLSA model, the paper accurately identified the sub topics and the feature words. Based on the text-words matrixes which had been ex-tracted before, this paper constructed the sentiment vocabularies. The application of similarity algorithm HowNet helped us to calculate the sentiment value of each word from the semantic angle, being positive or negative. Taking all of the feature words into consideration, this paper calculated the netizens' sentiments of each feature word in the network public topics.