摘要
为了改善电商虚假评论自动识别的效果,首先从传统的监督学习方法入手对网上商品评论的真实性进行判断,进而提出了利用社交图谱识别虚假评论。这种方法基于一种假设,就是同类用户通常在是否欺骗等行为上有相似性,将其结合传统的分类学习框架进行训练分类,实验结果显示社交图谱的方法能提高5%的识别准确率。
For the task of improving the performance of fake comments recognition in E-commerce, this paper firstly started with the traditional supervised learning methods to judge the truth of the online-commodities comments, then proposed the model of the social graph to identify fake comments. This method is based on an assumption that similar users usually have similar similarities on whether deception. The experimental results show that the social graph method can improve the recognition accuracy by 5%.
出处
《计算机应用》
CSCD
北大核心
2014年第A02期151-153,158,共4页
journal of Computer Applications
关键词
社交图谱
欺诈检测
监督学习
自然语言处理
语义识别
social graph
fraud detection
supervised learning
Natural Language Processing （NLP）
semantic recognition
