基于半监督学习算法的虚假评论识别研究

Deceptive Reviews Detection Based on Semi-supervised Learning Algorithm

已有的虚假评论识别方法主要采用启发式策略或简单特征建模。针对这些方法的不足,提出使用机器学习方法识别虚假评论。首先整合计算语言学与心理语言学的知识对评论文本进行建模,使用全监督学习算法来评价不同特征建模的性能,选出最好的特征组合。为了提高识别性能,设计2种半监督学习算法充分利用大量的未标注文本。实验结果证实所提算法超过当前的基准。

Machine learning methods were presented to identify deceptive reviews. With the integration of knowledge from computational linguistics and psycholinguistics,supervised method was developed to evaluate the performance of different feature modelings,and select the best mixed features. Then,two semi-supervised learning methods were designed to exploit the large amount of unlabeled data. The results showed the proposed methods outperform the current baselines.