Classifying Conversation In Digital Communication

Many studies of digital communication use natural language processing (NLP) to find topics, assess sentiment, and describe user behaviour. In finding topics often the relationships between users who participate in the topic are neglected. We propose a method of describing and classifying online conversations using only the structure of the underlying temporal network and not the text content of individual messages (using the recently proposed temporal event graph). The temporal event graph utilises all available information in the temporal network, i.e. no temporal aggregation of edges, and combines both topological and temporal structure using the widely-studied concepts of temporal motifs and inter-event times.
Considering the temporal event graph as a weighted network we can threshold the edges of the network to identify key timescales and decompose it into temporally overlapping components (or conversations). These components can be embedded into a feature space using a number of temporal and topological features. Furthermore, by performing a clustering of these components we are able to identify the particular behaviours of individuals and collectives in the temporal network and see how these behaviours evolve and combine over time. Finally, as this approach is context agnostic it can be used to understand other systems of temporally evolving behaviour such as protein interaction, human mobility, and financial transaction networks.

Andrew Mellor
Monday, September 24, 2018 - 18:00 to 18:15


The official Hotel of the Conference is
Makedonia Palace.

Conference Organiser: NBEvents

The official travel agency of the Conference is: Air Maritime

Photo of Thessaloniki seafront courtesy of Juli Bellou
fb flickr flickr