标题： 用户兴趣和在线论坛中的互动结构 显示英文标题

标题： User Interest and Interaction Structure in Online Forums

摘要： 我们提出了一种针对在线论坛中帖子的新型相似性度量。 我们的度量方法考虑了关于用户兴趣和互动的所有可用信息——帖子的内容、论坛中的主题线以及帖子的作者。 我们使用这种帖子相似性来建立基于主坐标分析的用户间相似性。 这使得用户活动的可视化变得简单。 用户间的相似性有许多应用，例如聚类或分类。 我们表明，在帖子相似性中包含帖子的作者会对主坐标投影产生平滑效果。 我们在从内部企业论坛中提取的真实数据上展示了我们的方法，并将结果与标准文档分类方法的结果进行了比较。 我们得出结论，我们的方法能更详细地描绘出局部和全局网络结构。 显示英文摘要

摘要： We present a new similarity measure tailored to posts in an online forum. Our measure takes into account all the available information about user interest and interaction --- the content of posts, the threads in the forum, and the author of the posts. We use this post similarity to build a similarity between users, based on principal coordinate analysis. This allows easy visualization of the user activity as well. Similarity between users has numerous applications, such as clustering or classification. We show that including the author of a post in the post similarity has a smoothing effect on principal coordinate projections. We demonstrate our method on real data drawn from an internal corporate forum, and compare our results to those given by a standard document classification method. We conclude our method gives a more detailed picture of both the local and global network structure.