提取文本流主题的神经网络新算法

目前，关于动态文本数据处理已逐渐成为数据挖掘的研究热点，例如，在聊天室中提取热门主题以及所有的讨论主题。目前已有的神经网络方法能较好地提取所讨论的主题，但不能决定哪个主题是热门主题，而且，提取到的主题之间相互干扰。利用主题之间相互独立和主题自相关的特性，基于自相关矩阵以及独立主元分析教学模型，本文提出一种新的神经网络方法，该算法能成功解决这些问题。在Yahoo聊天室上的实验结果表明，本文算法能准确提取主题以及热门主题，并且主题之间相互干扰大大减小。关键词独立主元分析，神经网络，自相关矩阵，时间序列

A New Neural Network to Extract Topics in Dynamical Text

Recently, the analysis of dynamically evolving textual data has become to an active research field in Data Mining. For example, extracting topics in the Internet chat lines. The existing neural network methods are based on linear time-serles model, which could extract topics very well. But it cannot decide which topic is the hot topic and the topics disturb each other. Since the topics is independent each other and the topics are self- correlation, a new neural network is derived. It can solve the mentioned problems. Simulation results on Yahoo chat room illustrate that our neural network indeed extract meaningful and hot topics. And the disturbance between topics is very small.