基于信息论的文本分类模型

Text classification model based on information theory

从信息论的角度,提出了一种新的文本分类模型。该模型以文本提供的关于类别的信息作为分类依据,从另一个角度来思考文本分类问题。从实用性的角度来看,该模型与传统的朴素贝叶斯模型和基于KL距离的中心向量法具有一定的关系,并给出了证明。根据广义信息论的基本概念,又对此模型进行推广,提出了特征权重的概念,可以通过修正特征权重来修正文本分类模型,为成功解决文本分类模型的修正问题提供了理论基础。

A new text classification model from the perspective of information theory is proposed. Considering text classification problem from another angle, this model employed the category information obtained from the text as the basis for classification. From the view of practicability, we proved it that this model has some relationships with the traditional naive Bayesian model and KL-distance based central vector method. According to the basic concept of generalized information theory, the promotion is carried on to this model and introduced the concept of feature weight, which has provided a foundational theory for solving the text classification model revision question successfully.