摘要： 信息时代作者版权的保护问题已受到越来越多的关注。针对部分小说存在的文本大面积相似问题，提出基于语言模型和特征分类的方法。统计文本二元~六元的语言模型并且绘制拓扑图，通过计算重合概率和词性比来分析词语的重合程度和语法信息，在此基础上利用主成分分析和随机森林的方法，进行分类判别。机器学习的结果表明，该方法能够有效地鉴别小说是否存在抄袭现象。

关键词: 抄袭判定, 语言模型, 语法信息, 主成分分析, 随机森林, 分类

Abstract: The protection of copyright property arouses much attention in the present information age. Aiming at the dispute problem caused by the text similarity between some novels, this paper proposes a method based on language model and feature classification, with statistics of coincidences and the proportion of pos to analyze the grammatical collocations and the coincidences. The methods of Principal Component Analysis(PCA) and Random Forest(RF) used to extract characteristics for automatic classification are added into experiments. The result of machine learning shows that the method can effectively identify whether novels exist plagiarism phenomenon.

Key words: plagiarism judgment, language model, grammatical information, Principal Component Analysis(PCA), random forest, classification

中图分类号: 

• TP311