Mining Frequent Patterns from Human Interactions in Meetings Using Directed Acyclic Graphs

In modern life, interactions between human beings frequently occur in meetings, where topics are discussed. Semantic knowledge of meetings can be revealed by discovering interaction patterns from these meetings. An existing method mines interaction patterns from meetings using tree structures. However, such a tree-based method may not capture all kinds of triggering relations between interactions, and it may not distinguish a participant of a certain rank from another participant of a different rank in a meeting. Hence, the tree-based method may not be able to find all interaction patterns such as those about correlated interaction. In this paper, we propose to mine interaction patterns from meetings using an alternative data structure—namely, a

directed acyclic graph

(

DAG

). Specifically, a DAG captures both temporal and triggering relations between interactions in meetings. Moreover, to distinguish one participant of a certain rank from another, we assign weights to nodes in the DAG. As such, a meeting can be modeled as a weighted DAG, from which weighted frequent interaction patterns can be discovered. Experimental results showed the effectiveness of our proposed DAG-based method for mining interaction patterns from meetings.