在线群体交互的影响模型及其学习算法

在线群体交互有助于数字图书馆发挥其服务人类需求的潜力，但如何量化在线群体交互对个人信息访问方面的影响还有待进一步研究．该文用隐Markov模型(HMM)来建模交互用户的状态序列及其相应的信息搜索行为，并基于影响模型理论提出了一个在线群体交互影响模型来分析用户在从数字图书馆选择资料和搜索所需信息时的相互影响．为满足本应用问题中增量模型学习的需要，文章还从耦合隐Markov模型(CHMM)学习算法引申出基于梯度的方法来进行在线群体交互影响模型参数的训练．实验结果显示，本文所提出的模型和算法能较准确地刻画在线群体交互对个人信息访问行动的影响．

The Influence Model of Online Social Interactions and its Learning Algorithms

This paper models the sequential states of each interaction participant and their corresponding information seeking actions with a Hidden Markov Model (HMM), and then proposes a common framework with the Influence Model to analyze the dynamic inter-influencing relationships of all participants on selecting material or seeking desired information from digital libraries. Furthermore, an incremental gradient-based algorithm is presented in this paper for the training problem of the above model. The experimental results show that authors' model and its learning algorithm can effectively characterize the influences people have on each other during online social interactions.