中文摘要：

为解决现阶段文本情感摘要方法缺乏的问题,提出一种基于改进情感关键句抽取的情感摘要生成技术。利用TFISF抽取文章的关键词,分析句子包含关键词数量,采用情感词典对句子进行情感程度分析,通过余弦相似度考察句子与主题的相关性,实现情感关键句的抽取并应用到情感摘要中。实验结果表明,该技术考虑了句子中的关键词属性、情感属性以及主题相关性,能够提高情感摘要的准确率。

英文摘要：

To make up for the lack of the sentiment abstracts technology,a method for sentiment abstracts extraction based on improved extraction algorithm of sentiment key sentences was put forward.To improve the method of extracting sentiment key sentences,the key words in the article were extracted using TF-ISF and the number of keywords was calculated.Emotional dictionary was utilized to analyze the sentiment orientation and the correlation of the sentence with the theme was explored using cosine similarity.The sentiment key sentences were extracted and applied to the sentiment abstracts.Experimental results show that the proposed method takes full account of the attributes of key words in the sentence,sentiment attributes,and the theme correlation.The accuracy of sentiment abstracts extraction is then greatly improved.