Extractive Summarization and Dialogue Act Modeling on Email Threads: An Integrated Probabilistic Approach
|Tatsuro Oya and Giuseppe Carenini|
In this paper, we present a novel supervised approach to the problem of summarizing email conversations and modeling dialogue acts. We assume that there is a relationship between dialogue acts and important sentences. Based on this assumption, we introduce a sequential graphical model approach which simultaneously summarizes email conversation and models dialogue acts. We compare our model with sequential and non-sequential models, which independently conduct the tasks of extractive summarization and dialogue act modeling. An empirical evaluation shows that our approach significantly outperforms all baselines in classifying correct summary sentences without losing performance on dialogue act modeling task.