摘要： 提出一种新的写作风格相似度评估方法，利用不同作者写作时在文章语句节奏控制方面的特点，鉴别作者的写作风格，从而达到作者身份识别的目的。该方法构建节奏特征矩阵模型来描述文本的语句节奏，利用点积相似度算法以及改进的KL距离算法来度量节奏特征矩阵之间的差异。实验表明，该方法在文学作品的作者识别方面具有较高的准确率。

关键词: 文本挖掘, 作者身份识别, 文本相似度, 节奏特征, 多维矩阵

Abstract: This paper proposes a new method of authorship similarity assessment, which identifies the authorship by sentence rhythm features of articles. The method constructs a rhythm feature matrix to describe the Sentence Rhythm Feature(SRF) of the text, and uses the inner product similarity algorithm and improves Kullback-Leibler(KL) divergence algorithm to measure the difference between the rhythm feature matrixes. Experiments show that it can make rather good results in literature authorship identification.

Key words: text mining, authorship identification, text similarity, rhythm feature, multi-dimensional matrix

中图分类号: 

• TP311.13