An Overview of a System for Relating Subjects in Meeting Minutes based on Community Structure

This paper describes an overview of a system for relating subjects in minutes of consecutive meetings. Meeting minutes are a collection of subjects. Moreover, subjects of a previous meeting may continue and may be discussed at a subsequent meeting, or similar subject matters may be discussed during meetings. Identification of similar subjects in minutes of consecutive meetings is necessary. Common named entities between subjects in different minutes appear to discuss similar matters. Therefore, the same named entities appear in related subjects, such named entities are organized into word communities as sets of named entities. Two subjects in minutes of consecutive meetings are related through word communities. The system provides mechanisms for constructing word communities based on a co-occurrence graph of named entities, and a topic model. There are three types of word communities, word communities based on modularity, a k-clique, and a topic model. A word community is a collection of related named entities that frequently co-occur. Because these mechanisms construct different communities based on different methods, the results obtained from the computed communities differ. Results can be compared using the relationships between subjects of minutes in consecutive meetings.