基于改进RI方法的文本聚类

The text clustering based on improved RI method

旨在通过考虑特征词汇的潜在语义和自身的重要性来提高文本聚类效果,研究基于RI方法的文本向量表示方法。首先,对基于RI方法构建的特征词汇随机索引向量中+1和-1向量元素出现位置进行约束,以避免在构建特征词汇上下文向量时可能造成该特征词汇潜在语义丢失现象;其次,在生成文本向量时考虑特征词汇自身重要性来改进权值的计算;最后,在测试数据上对基于RI方法的文本向量表示进行聚类效果测试与对比分析,结果表明采用基于RI方法能提高文本聚类效果。

In order to improve the effect of text clustering by considering the underlying semantic meaning and the importance of distinctive phrases in the text, research to take a method as text vector expression ways based on RI method. At first, constraint the position of vector elements ＋ 1 and -1 in a random index vector of distinctive phrase constructed by RI method, to avoid losing its underlying semantic meaning when construct the context vector of distinctive phrase; then,improve weighted calculation through consid- ering the importance of distinctive phrase when generate text vector; at last, testing and comparative analy- sis with the effect of text clustering by the method proposed in this paper of text vector construction on test data,and the test result shows that the method based on RI method could improve the effect of text clustering.