Experiments with Dynamic Topic Models
Jinjing Li, Wray Buntine · 2014
General summary of news content is a task under the general heading of“information summarisation, ” and it is recognised as a way of overcoming information overload. How does one summarise a large number of articles with their time as well as topical content? Here we introduce one technique, dynamic topic models, built using discrete non-parametric techniques, and demonstrate our software that can do this relatively efficiently using multi-core methods. Examples are used from a 760k collection of news articles from the Australian Broadcasting Commission (ABC) website over a ten year period.