独立于语种的文本分类方法

本文提出了一种独立于语种不需分词的文本分类方法。与传统文本分类模型相比，该方法在字的级别上利用了n元语法模型，文本分类时无需进行分词，并且避免了特征选择和大量预处理过程。我们系统地研究了模型中的关键因素以及它们对分类结果的影响，并详细介绍了评价方法。该文本分类方法已经在中文和英文两个语种上得到实现，并获得了较好的分类性能。

Language-Independent Text Categorization

The paper proposes an approach to language independent text classification without word segmentation. Unlike the case of traditional text classification models, the approach based on the character-level n-gram language modeling avoids word segmentation, explicit feature selection and extensive pre-processing. We systematically study the key factors in language modeling and their influence on classification, and describe an evaluation method in detail. Experimental results show that the proposed method can achieve good performance in text classification tasks.