摘要
文中提出了两个确定褒贬分类的最佳分界点划分方法,使用了一个多层的神经网络进行文本褒贬度的计算并对其输出值的分布进行分析。通过实验验证了这两个方法在褒贬分类中的有效性,并和朴素贝叶斯方法进行比较。实验结果表明,通过该算法确定的褒贬分界点较传统的分界点0.5具有较高的准确率。
The papers proposes two methods tO determine the best turning point between positive and negative sentiment in text sentiment analysis. A multi-layered neural network is employed to compute the value of sentiment. The two methods are verified through the experiments show that the turning point got by the methods has the precision comparable to that using default turning point. The comparison is also made with the traditional Naive Bayesian The experimental results show the improved precision.
出处
《江南大学学报（自然科学版）》
CAS
2009年第5期509-512,共4页
Joural of Jiangnan University (Natural Science Edition)
基金
国家自然科学基金项目(60673043)
关键词
文本分类
神经网络
情感分析
分界点
text categorization, neural network, sentiment analysis, turning point The results 0. 5 as the Classifier.
