时间信息在话题检测中的应用研究

Using Temporal Information in Topic Detection

为了克服话题检测中使用静态阈值的缺点,我们提出了基于时间信息的动态阈值模型。在该模型中,探索了一种比值法来选择与某个特定报道最相似的话题。实验结果表明,动态阈值模型很好地改善了话题检测系统的性能。

In order to overcome the shortcoming of the static threshold in the topic detection research, we propose a dynamic threshold model incorporating temporal information as a major component. In this model, we explore a ratio method to select the optimal topic. Experimental results indicate that the model proposed in this paper is very successful.