针对问答系统隐性垃圾内容的答案再排序模型

Re-ranking model for implicit spam answers in CQA

社会化问答系统为人们提供知识共享的平台,然而网站上存在着诸如推广信息的隐性垃圾内容,这些内容在包含诸多有用内容的基础上含有虚假的推广信息,这些虚假信息可能会带来更严重的后果,因此,如何检测及识别这些隐性的垃圾内容尤为重要。通过在任务型的网上交易平台搜集实验数据,创新地提出了一种基于物理学牛顿第二运动定律的优化答案排序模型,旨在原有的答案序列的基础上,加入隐性垃圾内容的特征,通过将回答者提交的答案看成是受多个力作用的物体,答案的排序看成是物体的下落过程,来对答案进行重新排序,使虚假信息沉淀到答案序列下方。实验证明,此模型能够快速有效地完成对答案的排序,实现按照质量对答案进行排序。

Community question answering system provides a platform for people to share knowledge. However, there are some content like the promotion of information hidden in the answers. It may lead to serious consequences. Therefore, how to detect and identify this hidden spam is particularly important. By collecting experimental data on a task-based online trading plat- form, this paper proposed an optimization model for re-ranking answers using physics of Newton＇ s second law innovatively. Each answer was seen as a falling object with several forces. And the answers would be rearranged, letting the spam informa- tion to the bottom of the answer sequence. Experiments show that this model can be completed quickly and efficiently in re- ranking the answer sequence.