ComModeler: Topic Modeling Using Community Detection

Loading...
Thumbnail Image
Date
2018
Journal Title
Journal ISSN
Volume Title
Publisher
The Eurographics Association
Abstract
This paper introduces ComModeler, a novel approach for topic modeling using community finding in dynamic networks. Our algorithm first extracts the terms/keywords, formulates a network of collocated terms, then refines the network based on various features (such as term/relationship frequency, sudden changes in their frequency time series, or vertex betweenness centrality) to reveal the structures/communities in dynamic social networks. These communities correspond to different hidden topics in the input text documents. Although initially motivated to analyze text documents, we soon realized the ComModeler has more general implications for other application domains. We demonstrate the ComModeler on several real-world datasets, including the IEEE VIS publications from 1990 to 2016, together with collocated phrases obtained from various political blogs.
Description

        
@inproceedings{
10.2312:eurova.20181104
, booktitle = {
EuroVis Workshop on Visual Analytics (EuroVA)
}, editor = {
Christian Tominski and Tatiana von Landesberger
}, title = {{
ComModeler: Topic Modeling Using Community Detection
}}, author = {
Dang, Tommy
and
Nguyen, Vinh The
}, year = {
2018
}, publisher = {
The Eurographics Association
}, ISBN = {
978-3-03868-064-2
}, DOI = {
10.2312/eurova.20181104
} }
Citation
Collections